|
SQ Leaderboard by Region
October 29th, 2011 05:39 GMT
Back in August, I was kicking back watching some of the top Korean pros battling it out on the MLG Raleigh stream. It's such a pleasure to see SC2 becoming an international esport, and it warms my heart to see the competition between Koreans and foreigners at MLG. It warms my heart even more to see foreigners winning over Koreans - go HuK! However, as I watched throughout the final day of MLG Raleigh (a.k.a. battle of the Koreans), something seemed amiss. First, Puma seemed to be getting supply blocked multiple times in every game. Then, NaDa started to look sloppy; in one game he was floating 1600 minerals while on one base! Hurricane Irene was passing through the area, and the amount of lag was getting serious, so it seemed only natural to attribute these slips to the less than ideal playing environment.
However, even taking that into account, I was a little shocked. From regularly watching GSL I felt that the top Koreans usually macroed much better than this. And how could NaDa, the macro-God, ever be floating that much money? His macro skills are off-the-charts! But what "charts" am I talking about? How would one actually go about proving that a player is macroing better or worse than they usually are? And if one had a reliable way of assessing that, wouldn't it be interesting to see how different players and regions of the world compare. At the time I was stumped; there didn't seem to be any straightforward way of measuring how well a player was performing in this respect.
Fast forward a few weeks, and I had just completed a large study of macro across the North American ladder. One of the outcomes of this analysis was a scale called the Spending Quotient (SQ), which measures numerically how well a player performed in terms of keeping their unspent resources low relative to their income (resource collection rate). Of course, this was not a comprehensive measure of macro, but it seemed to be the type of metric I was looking for.
Many readers suggested that it would be interesting to use this scale to compare pros from different ladder regions, and to see just how insanely the world's best players score on this scale. So, that's exactly what I set about doing!
This time around, I was fortunate enough to have some help in collecting data. Kudos goes to Flew for his superb work collecting data for Europe.
To begin the analysis, we collected ladder data for top level players on the American ladder. First, the order of the top 10 players in Grandmaster league was recorded. For each of these players, we then viewed their last 30 ladder 1v1 games. For most pros, this covered a time period of days to weeks. From each game, the following data were collected:
• Average Unspent Resources • Average Income (Resource Collection Rate)
We then computed the SQ for each of the 300 collected games, and calculated an average SQ score for each player.
This procedure was then repeated for the European ladder, and for other well-known players on both the American and European ladders who were not in the top 10 at the time of data collection.
We would very much like to extend this analysis to include players on the Korean, Chinese and South East Asian servers. If you are interested in contributing to this analysis and have an account on one of these servers, please PM me for more details. The names of contributors will be added to the article's header.
An important first question is whether SQ is even an appropriate scale for comparing top level players. Since SQ is only a measure of spending efficiency, it is quite possible that increased harassment results in lower SQ scores. Since higher level players are better at both macro and harassment, it is feasible that the two effects cancel out, meaning SQ can no longer distinguish differences in spending skill at the highest levels.
To investigate this, I tested whether the top 10 players on the NA ladder score significantly better SQs than the top 100 players. Previously, I had collected the last 3 games from each of the top 100 players on the American ladder, yielding a total of 300 games. Here, I compared the SQ scores for those 300 games to the 300 games collected from the top 10 players.
As can be seen above, the top 10 players score significantly higher SQ scores than the top 100 players on average (p<0.0001, two-tailed t-test). Particularly impressive is the fact that the gap between the top 10 (avg. SQ = 88) and the top 100 (avg. SQ = 82), is almost as large as the gap between the top 100 and masters (avg. SQ = 72). This finding is encouraging, since it shows that even at the top level, SQ continues to correlate with overall skill. Also interesting to note is the fact that there is smaller variation in SQ at higher skill levels; SQ standard deviation is 12 for masters, 11 for GM top 100, and 9 for GM top 10.
Plotting unspent resources against income also showed robust differences between the top 10 and top 100 players. Both groups followed an approximately exponential relationship between unspent resources and income, but with different levels of spending efficiency. Here, green corresponds to Masters, red corresponds to top 100, and blue corresponds to top 10.
Next, I tabulated the data from all the American accounts for which data were collected (including the top 10 and the other notable pros). The players were then ordered by average SQ score over their last 30 games.
Note that these are all extremely high average scores. However, it is interesting to note the amount of variability, even at the very highest levels of play.
The same analysis was performed for the European top 10, and other notable pros. The results are tabulated below, with players ordered by average SQ score over their last 30 games.
Coming soon.
Coming soon.
Coming soon.
Before getting into the analysis, there are numerous caveats that must be considered.
1) While SQ generally increases with skill, it is possible that different playstyles may lead to higher or lower SQ scores. And it may be that different playstyles are more common in particular regions. For example, let's suppose two base all-ins result in a higher SQ score, and two base all-ins are more common on a particular server. The average SQ score may be inflated for that region, despite the players not being inherently any better at spending. I have not found any correlations between SQ and game duration in any of the leagues, but that does not entirely rule out this possibility.
2) In my previous analysis, I found that each race scored equally well in terms of SQ at all levels of the ladder. However, it is possible that at the absolute highest level, the races run into different ceilings on spending. This would introduce a bias against certain players, and may even slightly affect the average SQ for Grandmaster league between regions if there are different racial compositions. I don't yet have enough data to test this hypothesis.
3) If SQ is indeed balanced for all races, it is nonetheless possible that different match ups could differentially affect SQ. For example, Protoss could score the same as the other races on average, but score worse in PvP and better in PvT and PvZ. This is an analysis that I've not yet performed, so I don't know whether this is the case or not.
4) It's important to remember that many of the top players in each region are frequently in other regions (e.g., Korea). Furthermore, players often play on multiple regions. This means the top 10 for a particular region is not necessarily representative of the depth of the skill pool in each region. Caution needs to be exercised in analyzing the results - they should not be used to conclude that players from region A are better than players from region B.
5) It's been said before, but I'll say it again: SQ is not everything. The Spending Quotient does not take any strategic factors into account. It simply measures how well a player spends their money relative to all other players at the same level of income. When averaged across many games, it has been shown to correlate well with overall skill, and it measures one of the core skills. But it does not directly measure anything else.
With those points all in mind, let's go ahead. To compare the spending performances for the top players between ladder regions, I plotted the distributions of SQ scores for the top 10 ladder-ranked players in each region.
As you can see, the distributions for the American and European servers are actually very similar. However, the average SQ is slightly higher for America than Europe (88 vs. 84), and the difference is statistically significant (p<0.0001m two-tailed t-test).
Two players appear on both the European and American servers. These are ToD (currently based in Europe) and EGDeMusliM (currently based in the US). These players provide a unique opportunity to test whether SQ is robust between servers, especially when playing on a non-local server (presumably with slightly higher lag).
In both cases, the players score very similarly on both servers. Neither difference is found to be statistically significant (p = 0.4 for EGDeMusliM; p = 0.9 for ToD). EGDeMusliM's standard deviation is a little larger for the American games than for the European games; this may just be a quirk of the 30 game sample.
Below is the leaderboard for all pro players that have been analyzed to date with an average SQ greater than 80. Colors correspond to regions, and some players appear on multiple servers.
EGDeMusliM, who appears near the top of this list on two servers, seems to be almost as good at spending as he is at breaking his arm.
Comparing SQ scores on the ladder is one thing. But ideally, one would want to compare pros playing in the exact same environment, such as a tournament. To address this, I also analyzed the SQ score for all top 8 players at MLG Orlando 2011. I did this by searching through the match histories of the players during the tournament - a different way of following the action live!
Below are the summary statistics for each player, presented as player cards.
This was a really nice opportunity to compare players from multiple regions, and it generated some interesting findings. Chiefly:
1) Pros are incredibly good at spending
Well, duh. But still, it's impressive to see just how good. Bomber scored 115 in one game at MLG. Try doing that! The ladder analysis turned up only two scores better than this: 116 by EGdeMusliM, and an astounding 121(!) by DroneKing (LiquidRet). For those interested, that last score corresponded to a Resource Collection Rate of 2239 and an Average Unspent Resources of just 653. Truly incredible efficiency.
2) Pros turn it up to 11 at tournaments
One might expect that pros macro even better in a tournament environment than on the ladder. This appears to be true, at least based on a comparison of IdrA's ladder and tournament performances:
The difference in means is statistically significant (p = 0.02, two-tailed t-test). It would be intriguing to extend this analysis to include other pros to see the gulf between their ladder and tournament performances. It may even be that nerves cause some players to macro worse at tournaments. In the case of IdrA, however, the ladder-Gracken doesn't seem to be at full capacity.
3) Protoss scored lower SQs at MLG
Among the top 8 players, Protoss scored notably lower than other races. While this is an extremely small sample of players, I found this surprising. Especially since this sample included two of the best Protoss players in the world (HuK and MC). Whether this down to their particular playstyles, skill sets, or some sort of systematic SQ bias against Protoss due to the race's mechanics will require additional data from other top Protoss players to test. It certainly seems possible for Protoss players to score extremely high SQs in individual games, but the highest average SQ observed for a Protoss player to date is 89 (ToD). If no higher SQs are found once the Korean ladder is analyzed, it may be necessary to adjust the way SQ is scored at the top level.
4) Spending ain't everything
Of the top 8 players, the lowest SQ belongs to the winning player, HuK! This shouldn't be so surprising, since there are clearly other factors that contribute to overall skill, and HuK is particularly good in the micro department. This doesn't mean SQ isn't a useful metric, but it does emphasize that it is not a comprehensive measure of skill - and we shouldn't expect it to be.
I hope you find this analysis interesting! I think it turned up a bunch of interesting findings, and again it confirms that SQ is a good metric for self-assessment. However, it also opens some interesting questions, including whether SQ is being fair to Protoss at the highest level. Since the production cycle works differently for different races, it is possible that the races run into different ceiling values of SQ at the very highest level. This doesn't seem to be an important factor for regular Grandmaster players down to Bronze, but for pros it may be important to consider.
I look forward to hearing feedback from the TL community. I'd be especially interested to hear whether people would be interested in seeing further analyses like this in the future, and if there are any suggestions. Finally, if anybody could help to collect data for the remaining servers, I would be very thankful.
|
amazing article man. put a lot into it. can't wait to see all the other regions!!! Great read and very impressed!!!
|
Wow...this is am impressive write-up and a ton of information. I read through most of it but it will take awhile to digest this all. Thanks!
|
I think the warpgate mechanic might actually be hindering Protoss' SQ, but very nice read. I'm pretty surprised at Bombers SQ and Ret's as well.
|
SQ must have a bias against protoss from looking at the leaderboards. NA has 4/16 and EU has 3/18, and all the protoss aren't very high.
|
I also analyzed the SQ score for all top 8 players at MLG Orlando 2011. I did this by searching through the match histories of the players during the tournament - a different way of following the action live! You are one smart dude. WTF, too good. Can't thank you enough.
|
Great data and write up. However, because protoss units typically come in waves through gateways would that affect the SQ at all? It's not rare for protoss to stack minerals so they can get all their units off the gateways at the same time.
|
On October 29 2011 10:50 MonkSEA wrote: I think the warpgate mechanic might actually be hindering Protoss' SQ, but very nice read. I'm pretty surprised at Bombers SQ and Ret's as well. That is what I suspect. Either that or something to do with the types of builds, e.g., it is beneficial to build multiple gateways all at once to sync up with warpgate research, meaning you have to temporarily float money. But I can't conclude that without more data.
|
|
Impressive write-up, will read it again to grasp some details that I skimmed over.
|
Great article.
Protoss can be explained by the fact that it is the only race where it's sometimes actually good to overmake warp gates and bank resources. I suspect that is why we see lower protoss SQ especially in tournaments; planned strategies probably take the into plan more than ladder strategies, and the very best protoss are more likely to implement this often.
|
out of interest, how do we find out our own SQ?
|
shit tod is the king of protoss macroers :[
|
On October 29 2011 11:01 unit wrote: out of interest, how do we find out our own SQ? The original article shows how to calculate this, and some people in the community have actually created tools to make this easier! Check out the FAQ at the bottom.
|
wow, this makes me think so much more of demuslim, and I was already high on him
|
United States7483 Posts
One issue I have:
Zergs, when going for muta builds, deliberately pool resources and float them for a while before the mutas come out, which raises their average unspent resources a fair bit. Going for infesters they float a lot of gas as well, etc.
Protoss play a heavy gateway style often will gain 600+ minerals and a bunch of gas too before their warp gates come off cooldown and then they burn them all at once. That would raise their average unspent resources.
Terran is the only race that doesn't float a lot of resources by design.
One should compare the metric between people playing the same race, but not compare across different races.
|
This is amazing, thanks. I love demuslime always have, who is faith? funny haven't heard much of him.
|
loving the sabermetrics of starcraft
|
On October 29 2011 10:40 whatthefat wrote:
Among the top 8 players, Protoss scored notably lower than other races. While this is an extremely small sample of players, I found this surprising. Especially since this sample included two of the best Protoss players in the world (HuK and MC). Whether this down to their particular playstyles, skill sets, or some sort of systematic SQ bias against Protoss due to the race's mechanics will require additional data from other top Protoss players to test. It certainly seems possible for Protoss players to score extremely high SQs in individual games, but the highest average SQ observed for a Protoss player to date is 89 (ToD). If no higher SQs are found once the Korean ladder is analyzed, it may be necessary to adjust the way SQ is scored at the top level.
In addition to Pokebunny's comment above about Protoss intentionally overmaking gates and banking resources, it should be noted that there is a natural buildup of resources during each warpgate cooldown, resulting in a lower SQ even were the Protoss to not overmake gates.
I also would like to point out that this measures only the spending of money, not the efficacy of that spending. As far as I understand it, a player poorly macroing and queuing up several units would receive a high SQ because their money was being spent, even if it was not being done in a productive way.
Edit: misquote
|
Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units.
|
impressive!!! cant wait to see all the rankings and results
|
[quote]I hope you find this analysis interesting! [quote] Oh hell yes. More stuff like this, please! While I'm not quite sure how to interpret the data, it is intriguing - especially the protoss bit.
Very well-written and argued, with some very healthy skepticism towards drawing any conclusions at this stage. AAAAAAA++++++ would read again
|
On October 29 2011 11:13 Imperium11 wrote:Show nested quote +On October 29 2011 10:40 whatthefat wrote:
Among the top 8 players, Protoss scored notably lower than other races. While this is an extremely small sample of players, I found this surprising. Especially since this sample included two of the best Protoss players in the world (HuK and MC). Whether this down to their particular playstyles, skill sets, or some sort of systematic SQ bias against Protoss due to the race's mechanics will require additional data from other top Protoss players to test. It certainly seems possible for Protoss players to score extremely high SQs in individual games, but the highest average SQ observed for a Protoss player to date is 89 (ToD). If no higher SQs are found once the Korean ladder is analyzed, it may be necessary to adjust the way SQ is scored at the top level. In addition to Pokebunny's comment above about Protoss intentionally overmaking gates and banking resources, it should be noted that there is a natural buildup of resources during each warpgate cooldown, resulting in a lower SQ even were the Protoss to not overmake gates. I also would like to point out that this measures only the spending of money, not the efficacy of that spending. As far as I understand it, a player poorly macroing and queuing up several units would receive a high SQ because their money was being spent, even if it was not being done in a productive way. Edit: misquote Well, let's remember that all the races score equally well on SQ for Bronze through Grandmaster (top 100). It's only at this extremely high level that I'm seeing any racial differences (and they aren't yet confirmed).
And yes, there are ways to 'game' SQ, e.g., building way too many production facilities for your income, or queuing like crazy. But even still, on average Bronze players score worse than Silver players, who score worse than Gold players, and so on. So it still seems to be a pretty good metric for assessing your spending ability. But it shouldn't be viewed in complete isolation - of course if you are doing something else terrible, like queuing all the time, you need to address that. I don't think any pro players are doing that though.
Once we have more data we can figure out whether the SQ needs to be slightly adjusted for race at the pro level - or just compared within races as Whitewing suggested.
|
kinda odd seeing Goody up there, I haven't kept up with him, but he's known for queuing the shit out of his factories right? I guess that doesn't really factor in SQ, and if he would just build more factories then his macro and relative SQ would make sense
|
Hmm I don't know if this was addressed but does the SQ rating account for those long-ass macro games where you're already maxed and there's nothing to spend your money on so you end the game at X000m, X000g?
|
Protoss is of course going to have higher unspent, zergs and terrans can make 50 mineral units. So all game long a protoss player would have to wait for at least 100 before he can spend it. Not only that but because of warp cycles, protoss is always going to be stacking up minerals in between cycles. Where a zerg can constantly spend if he has larva and a terran can cue units. As a game goes on there are going to be more gates added and more accumulated from income in between warp cycles. All of this artificially inflates a protoss players average unspent resources. Especially when you look at the cost of stalkers, colossus, immortals compared to anything terran or zerg are producing in the mid game. Protoss has to let more pile up before it gets spent. On top of that most protoss upgrades and tech structures are 2x the cost of terran or zergs.
This clearly demonstrates protoss having more unspent. I think it is purely a result of the race mechanics. It could probably be factored out by subtracting a certain 10 factor of unspent minerals from protoss. If you look at enough averages I'm pretty sure you could accurately calculate what the amount subtracted should be.
Edit: My bad thought that was by race. Here it is, it's still close but there are a lot of green dots above the rest. That could be a reflection of skill or race mechanics. You can almost see on this chart when warp finishes the toss unspent jumps. OP is obviously going to be 10x better at drawing an accurate conclusion than I. But excellent post! and thank you!
Also for consideration is that once a zerg maxes they are going to stack like crazy. So @ 200 supply the unspent really shouldn't be counted. (if it is)
|
Katowice25012 Posts
Everything you post is amazing, great stuff
SPOTLIGHT
|
On October 29 2011 11:21 whatthefat wrote:
Well, let's remember that all the races score equally well on SQ for Bronze through Grandmaster (top 100). It's only at this extremely high level that I'm seeing any racial differences (and they aren't yet confirmed).
And yes, there are ways to 'game' SQ, e.g., building way too many production facilities for your income, or queuing like crazy. But even still, on average Bronze players score worse than Silver players, who score worse than Gold players, and so on. So it still seems to be a pretty good metric for assessing your spending ability. But it shouldn't be viewed in complete isolation - of course if you are doing something else terrible, like queuing all the time, you need to address that. I don't think any pro players are doing that though.
Once we have more data we can figure out whether the SQ needs to be slightly adjusted for race at the pro level - or just compared within races as Whitewing suggested.
These are some fair points. I agree that on the whole SQ seems to be a decent metric of skill measurement, as your correlation with increasing league level shows. I just also think it is good to keep these inherent flaws in mind as they can explain unexpected variance (such as already described above regarding Goody).
|
So i took the the time to calculate my SQ and found to things
1. My macro is alot better then i thought it was with a SQ of 80 over 15 games
2. I macro like a beat when i proxy gate =P average proxy gate SQ 92
|
On October 29 2011 11:35 Reborn8u wrote:Protoss is of course going to have higher unspent, zergs and terrans can make 50 mineral units. So all game long a protoss player would have to wait for at least 100 before he can spend it. Not only that but because of warp cycles, protoss is always going to be stacking up minerals in between cycles. Where a zerg can constantly spend if he has larva and a terran can cue units. All of this artificially inflates a protoss players average unspent resources. Especially when you look at the cost of stalkers, colossus, immortals compared to anything terran or zerg are producing in the mid game. Protoss has to let more pile up before it gets spent. On top of that most protoss upgrades and tech structures are 2x the cost of terran or zergs. This clearly demonstrates protoss having more unspent. I think it is purely a result of the race mechanics. + Show Spoiler +
Please note that the graph colors correspond to the graph above that, i.e., green = masters, red = top 100, blue = top 10. It's not showing races - they end up pretty much on top of each other (you can actually see the top 100 plot by race in the original SQ thread). Sorry, I should have made that clear in the OP - I'll edit now.
|
On October 29 2011 11:35 heyoka wrote: Everything you post is amazing, great stuff
SPOTLIGHT <3 Thanks!
|
|
This guy has put SO much time into his posts. Kudos
|
these things are always interesting and yet depressing to me at the same time.
when averaging games with builds ive refined and feel comfortable with my sq averages at 92 lol... and ive been stuck in master =/
On October 29 2011 11:01 Pokebunny wrote: Great article.
Protoss can be explained by the fact that it is the only race where it's sometimes actually good to overmake warp gates and bank resources. I suspect that is why we see lower protoss SQ especially in tournaments; planned strategies probably take the into plan more than ladder strategies, and the very best protoss are more likely to implement this often.
and then this is something ive actually changed instead of spending it instantly ive been trying to bank a little more and make better decisions, my sq dropped and my MMR went up to the point i got paired with you lol...though its awkward yet i think your so right. i hope the next time i get you on the ladder i dont choke =P
|
I read this. I'm afraid I don't understand most of the statistics, but those results sure are interesting to look at. Great work!
|
United States11933 Posts
|
Very interesting and detailed analysis.
|
On October 29 2011 10:50 MonkSEA wrote: I think the warpgate mechanic might actually be hindering Protoss' SQ, but very nice read. I'm pretty surprised at Bombers SQ and Ret's as well.
this. a lot of players like to use all their warp gates at once to "save apm" so during the cool down the resources pile up.
|
This does give inspiration to blizzard to include in the end game stats total queue time of units.
For example, if there are 2 units queued up, not being built, than the Total Queue Time counter would be going up by 2 a second.
Then if Blizzard could also reveal game length in the statistics, you could make a similiar SQ, for queue time.
|
With the large amount of tournament replays available, it would be nice to put as many tournaments together into this as possible, perhaps segregated by month?
Additionally, there has to be a reasonable web application or tool which people can submit their own data into quickly, to increase the sample size significantly.
|
|
Awesome work as usual, I'm interested in seeing any differences between the KR server and NA and EU when that data is released as well.
|
On October 29 2011 11:57 GhostFall wrote: This does give inspiration to blizzard to include in the end game stats total queue time of units.
For example, if there are 2 units queued up, not being built, than the Total Queue Time counter would be going up by 2 a second.
Then if Blizzard could also reveal game length in the statistics, you could make a similiar SQ, for queue time.
This wouldn't work at all for zerg, unless it was possible to track unspent larva, but even then would be extremely difficult to match to the stats for protoss or terran. Warpgates would be similarly difficult.
|
|
I love these. Huk so good. Idra macro so good!
|
As expected, STC and Bomber with 2 of the highest SQ's! <333333
Also, it should be noted that RGNartist and Rebornsaga are the same people.
|
IdrA's best SQ was against Bomber on Shakuras, really showing his A game when it counted.
|
Is there any way to see how SQ changes with respect to game length? It seems to me that it was be easy to have a high SQ in a short game where you are only on 1-2 bases, but in drawn out games it becomes much tougher (supply cap). It would be interesting to see how well the pros do compared to other GM or Master level players in keeping their SQ high during long games.
|
On October 29 2011 12:13 Achilles306 wrote: Is there any way to see how SQ changes with respect to game length? It seems to me that it was be easy to have a high SQ in a short game where you are only on 1-2 bases, but in drawn out games it becomes much tougher (supply cap). It would be interesting to see how well the pros do compared to other GM or Master level players in keeping their SQ high during long games.
Here's an analysis I did a while ago with respect to game duration for the top 100 players. I didn't find any significant effect of game duration (and results were similar for lower leagues).
|
This is incredible work, very interesting! DeMusliM - what a baller.
|
I loved your first write up and was super glad to see you went on with this and did more with it.
With this and APM we might be closer to reaching the metric for the next best.
|
Very nice. I've long thought that SC2 analysis was in desperate lack of a rigorous statistical treatment and it's great to see you stepping up to the plate to provide it. Hopefully we'll see a sabermetrics-like community grow around SC2 in the near future, haha.
|
Canada13372 Posts
On October 29 2011 11:13 Imperium11 wrote:Show nested quote +On October 29 2011 10:40 whatthefat wrote:
Among the top 8 players, Protoss scored notably lower than other races. While this is an extremely small sample of players, I found this surprising. Especially since this sample included two of the best Protoss players in the world (HuK and MC). Whether this down to their particular playstyles, skill sets, or some sort of systematic SQ bias against Protoss due to the race's mechanics will require additional data from other top Protoss players to test. It certainly seems possible for Protoss players to score extremely high SQs in individual games, but the highest average SQ observed for a Protoss player to date is 89 (ToD). If no higher SQs are found once the Korean ladder is analyzed, it may be necessary to adjust the way SQ is scored at the top level. In addition to Pokebunny's comment above about Protoss intentionally overmaking gates and banking resources, it should be noted that there is a natural buildup of resources during each warpgate cooldown, resulting in a lower SQ even were the Protoss to not overmake gates. I also would like to point out that this measures only the spending of money, not the efficacy of that spending. As far as I understand it, a player poorly macroing and queuing up several units would receive a high SQ because their money was being spent, even if it was not being done in a productive way. Edit: misquote
All this makes Idra's score and the crazier. No queue and no reason to bank resources the same as protoss
|
Sweden33719 Posts
Your picture of thestc is actually oGsEnsnare :p
|
this might've been mentioned but that picture of oGs.TheSTC is actually of Ensnare. Otherwise my friend, you've done some amazing work and this is pretty awesome in general. I wonder if someday we'll be looking at statistics like that in tournaments/daily's(looking at you day9) to help quantify how to be a better player.
cheers
edit: jinro beat me to it, kid's got skills :p
|
Wow this is actually pretty sweet, both interesting and easy to grasp!
|
If you think about when people pool money, it makes sense that T should be highest SQ, then Z, then P.
T doesn't really have any reason to ever float money, other than preparing to expo for turrets + CC.
Z pools for tech switches (mutas, roaches, etc) and actively tries to pool once hitting 200/200 supply.
P constantly pools between warp cycles, so should just consistently be above the average of others.
I am actually crazy excited for this once we can compare the data to Korea! Keep up the awesome work.
|
On October 29 2011 12:31 Liquid`Jinro wrote: Your picture of thestc is actually oGsEnsnare :p Haha, good catch. Thanks - fixed!
|
Really amazing write up. Can't wait for you to get your hands on more raw data!
|
Well, aside from things like warp in and queue types of mechanics and micro, it's also obviously very important precisely what you are spending your money on, rather than simply how much of it you are spending. Huk is a great example of this. The main reason his spending quotient is so bad is not because he isn't capable of macroing properly; the guy practices like 12 hrs a day in Korea, I'm pretty sure he can handle protoss macro mechanics. I think the main reason Huk's spending quotient is so bad is because he's really really REALLY careful about what exactly he spends his money on. For example, while most players would just warp in whatever gateway unit composition fits their resource count best, Huk warps in the exact number of zealots, stalkers, and sentries respectively that he needs for each situation, even if his resources end up looking very uneven after the production cycle. For instance, as a frequent stream fan girl of Huk I've noticed that he would prefer warping in 5 zealots to 10 sentries if he felt the situation demanded zealots, even if he ended up floating 1000 gas temporarily because he only made zealots.
|
I'm actually surprised at how low the SQ is for the ladder players. That is terrible.
Maybe seperate coefficient needed for each race since protoss pool money and then mass warp in.
I think some of korea's highest SQ is going to be polt and MVP.
Nic ewriteup! curiosu about the rest of the data
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units.
Exactly what I was thinking.
Great analysis and really opens our newbie eyes on how extremely important spending and earning money is. While most people will say that is common sense, you probably/most likely did not understand the extremes. Thank you for the incredible analysis Whatthefat.
Interesting results between EU and NA I must say...
|
Yeah, Terrans should have the best SQ overall. Being able to queue your production facilities to the nth degree really helps sinking money
|
|
Very nice right up! You always have great posts.
|
You are such a baller whathtefat, this is really interesting . Although I do think it's hard to calculate for toss because of wg mechanic
|
this is really amazing
It would be great if this turned out to be a good predictor for tourny results but i think with SC2 this just won't be the case
I would bet though that this would be a decent predictor for BW
|
On October 29 2011 12:19 whatthefat wrote:Show nested quote +On October 29 2011 12:13 Achilles306 wrote: Is there any way to see how SQ changes with respect to game length? It seems to me that it was be easy to have a high SQ in a short game where you are only on 1-2 bases, but in drawn out games it becomes much tougher (supply cap). It would be interesting to see how well the pros do compared to other GM or Master level players in keeping their SQ high during long games.
Here's an analysis I did a while ago with respect to game duration for the top 100 players. I didn't find any significant effect of game duration (and results were similar for lower leagues).
Thanks! I agree that it doesn't seem to matter.
|
Another excellent statcraft thread - thanks. I've got a strong gut feeling that, as you mentioned, the warpgate mechanic (or rather, the way that lets players build gateways instead of units while deliberately floating money) has a lot to do with the lower Protoss scores, so it'll be good to have more data to analyze that with!
|
TLO has surprisingly good macro despite what people think about him! Very interesting.
|
Incredible post. I'm kind of a math junkie so could you go a little more into detail of what a number above 100 in the SQ means? I recall you writing on your original SQ post that numbers above 100 were extreme cases. Does the SQ take into account queuing units in a production facility, for example, 5 SCVs queued on a single CC? This would make spending money easier as Terran, since you have to wait for larva as Zerg and warp gate cycles as Protoss.
Still, congratulations on your analysis, very interesting read!
|
Very nice post. Great job grabbing that tourney data, fantastic idea.
|
|
really great work, thanks so much!
|
I also find it hard to understand why the SQs would be similar across races. I know from the first post on this subject, you said there seemed to be no statistical difference between races, but like many people here I find it hard to understand why it isn't easier to have better SQ as Terran because of queuing vs having Protoss's warp-in mechanic.
I'd definitely be interested in the SQ for All leagues in Korea, since a long standing debate is stuff like if Gold Korea > Platinum America etc...
|
Great job!
I do hope this topic will be re-spotlighted once the other regions are ready.
|
intrigue
Washington, D.C9931 Posts
you are an obscenely good poster
|
=drools= this is some next level statistical genius...thank you for all the glorious numbers
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. this. the data will be skewed a little =\
|
On October 29 2011 10:50 MonkSEA wrote: I think the warpgate mechanic might actually be hindering Protoss' SQ, but very nice read. I'm pretty surprised at Bombers SQ and Ret's as well.
I was thinking this aswell but it's 7am here so I couldn't wrap my head around why that is other than some generic thougths. If someone could explore this are a bit in this thread I would greatly appreciate it.
|
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units.
yea SQ rly doesnt mean much I feel kind of bad for the guy who did all this work. It's interresting but rly noobs should stop thinking of it as something more reliable than game score or apm
|
On October 29 2011 14:27 awesomo0O wrote:Show nested quote +On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. yea SQ rly doesnt mean much I feel kind of bad for the guy who did all this work. It's interresting but rly noobs should stop thinking of it as something more reliable than game score or apm
What? That is the complete opposite of what you are supposed to conclude from the analysis.
|
On October 29 2011 14:27 awesomo0O wrote:Show nested quote +On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. yea SQ rly doesnt mean much I feel kind of bad for the guy who did all this work. It's interresting but rly noobs should stop thinking of it as something more reliable than game score or apm
I'd say the ability to spend resources is generally of a lot more importance than apm and game score.
|
This is interesting. Thank you!
|
On October 29 2011 14:27 awesomo0O wrote:Show nested quote +On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. yea SQ rly doesnt mean much I feel kind of bad for the guy who did all this work. It's interresting but rly noobs should stop thinking of it as something more reliable than game score or apm
If you think how well you you spend your money is less important than game score or apm, you're a fool.
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units.
Yet ToD has one of the best SQ avg. (ranking 5th). I'm pretty sure it was concluded in the original article explaining SQ that there were no differences in SQ within the races.
|
Katowice25012 Posts
Whatthefat is now part of our TL writing staff, and as such we have made this featured news. Enjoy!
|
can you tell me how u calculated this? im trying to do the same thing for my own replays as a presentation. as of now im only using the avg unspent resources tab in economy breakdown after replay. please pm me
|
Yes, please keep doing these types of analyses. Your knack for scientific curiosity in ESports keeps me intrigued every time. Thanks for your hard work, and great job.
|
Katowice25012 Posts
On October 29 2011 14:42 Leeoku wrote: can you tell me how u calculated this? im trying to do the same thing for my own replays as a presentation. as of now im only using the avg unspent resources tab in economy breakdown after replay. please pm me
It's in his original post that he quotes within the first paragraph.
|
Fantastic post. Someone get this man data for the other servers ASAP.
|
Thank you for the effort you put into doing this ^_^
Always interesting to read these kinds of stats.
|
HuK won in part because of his lower SQ. Protoss benefits from floating money to be able to warp in extra units wherever they want and HuK has said in interviews that he often overmakes warpgates for this reason. It appears to also be strongly race dependent. All three Tosses had their highest SQ on PvZ match ups. This is probably because that matchup favours less warpgate units. In PvP, the 4warpgate is common. In PvT, money must be floated to defend from drops.
|
To the front page! Awesome to see admins supporting this effort.
|
Great post man, thank you very much
|
This is unbeleivable however the margin of error is still pretty significant because of what can happen in a game and that requires to save some minerals etc.
However, if all these players played eachother for 30 games and the study was conducted it may be different but thats being super picky.
regardless really really awesome read thanks alot man
|
Bookmarked. Can't wait to see the other regions.
|
This will only get better as more and more games and players are added into the mix. I could see this being something that in the future Blizzard would want to make possible to calculate on your SC2 profile if this gets enough steam behind it.
|
wouldnt the data be skewed from maxed out armies, because once you can't produce anymore you start to float minerals and such.
also, its kind of bad to compare each race to another, since each has different macro mechanics (i.e. larva, warpgates). a terran with unspent resources is a lot worse than a zerg or toss.
i can't code that well, but if you used a program like sc2gears or something that analyzes replays, you could get a lot more data a lot quicker. have it calculate the SQ of a player up until the player hits 195-200 supply. that seems like a better indicator of pure macro imo.
|
Such a great job on this, as always.
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. Yeah I was about to saw just that. If you have 7 gates as protoss with just enough money to use them you will like like 800 resources right before your cycle, so maybe the data is not 100% accurate but it is still a really cool thing you did. Really like the breakdown of the regions ect.
|
Ret is so good at macro. . . .
|
Zerg have to pool resources as well while waiting for larval inject to finish; in a sense, both zerg and protoss macro come in waves of spending, raising average unspent resources.
|
Awesome Sauce. EDIT: IF you made the SQs relative to other SQs from the same race, it would HAVE to mean something.
|
it's nice to see this SQ idea get more elaborated and investigated than it was before
Thanks a lot!
|
damn, DeMuslim is tearin it up
get it.. TERRAN it up? lulz. im so lame
|
always knew ret was a macro beast the very first game i saw of him where he had 4 queens with 5 energy each 17 minutes into a game.
|
Impressive research and awesome wright up!
Interested in the other servers. I want to see sen :D
|
Does SQ account for people that queue units?
For example, If I have 250 minerals and I build 5 marines in one barracks my money will be at 0 but really 200 of it could of been better spent elsewhere.
|
as you found out that your SQ is only one side of the coin (see Huk) maybe you could device a similar measure for control by using the units lost / resources lost data. Would probably require similar effort to normalize but would be interesting if you can analyze players in different categories. Might even be used in the predictor project to increase accuracy.
|
intrigue
Washington, D.C9931 Posts
On October 29 2011 15:25 d9mmdi wrote: as you found out that your SQ is only one side of the coin (see Huk) maybe you could device a similar measure for control by using the units lost / resources lost data. Would probably require similar effort to normalize but would be interesting if you can analyze players in different categories. Might even be used in the predictor project to increase accuracy. that's a really cool idea. it'd be the.... "using quotient" to complement the spending quotient? a standardized unit for unit efficiency would be fascinating to study.
|
Very interesting read. I look forward to any future writing on the subject.
|
After rereading this, it excited me a lot because it tells me that we are not anywhere near the skillcap for SC2 and games/players have a ton of room to improve.
I think back to hardcore macro in SC1/BW where the best players spend their money almost perfectly. Once each race and players evolve more, I imagine that the SQ will continue to go up. If people are not spending their money, then there is room to continue to grow strats and builds to maximize resource.
I love this. I would love to see this done for high-level Brood War as well. I know it is a different game, but it has a similar macro/econ game that is much closer to perfection at the top levels.
|
Why Demuslim and ToD have better SQ on foreign Server?
Demuslim has 92.6 SQ on America ,where is is acutally and 95SQ on Europe ToD is kinda the same 88.7SQ on his Home Server Europe and 88.9 on America.
I my self play on Korean often since 2 weeks. I my self recognized that I macro better on KR. This is because the lag do not let you micro so good. So you have to macro to compensate the lag of micro. Macro do not need a low ping. Everything in macro feels the same, it just happen 0,5 seconds later^^.
BTW I am on Europe^^
|
this post is amazing. and like a lot of previous posters, i think that warpgate probably has some factor at play here.
so impressed by this that i'm going to look up your first thread now! please keep it up!
|
Another great read, and useful as well as informative I might add, from the best website in the world, beating out Facebook, Twitter, and YouTube.
|
On October 29 2011 14:42 heyoka wrote: Whatthefat is now part of our TL writing staff, and as such we have made this featured news. Enjoy! Congradulations Whatthefat! I look forward to more of your articles.
|
Protoss having lower SQ doesnt surprise me because of having warp in rounds. I have also found that at every level of play I have been, atleast according to SC2gears, protoss have about 60% the apm of average terran and zergs. (notably gold through to diamond) I hope blizzard sorts out the chronoboost mechanic so the better protoss players can more easily differentiate themselves from the lesser ones. At the moment many protoss games feel like top trumps, he shows his hand and one of us wins, theres no give and take.
Thanks for the article, reminds me that macro differentiates players at all levels. I'd love it if people explored this more thoroughly though, the game sample seems rediculously small.
|
Suprising huk has one of the lowest at MLG, but he still kicked ass! great writeup
|
Awesone awesome article. that must have been a lot of work.
|
link to the 121 sq replay?
|
So I finished a very simple, and probably slightly buggy, interface to calculate SQ. Unfortunately there's no way for it to read and write on a singular file for long term use yet, but if there's enough demand I could feasibly make one, fixing most of the bugs in the mean time. Here's the link: http://www.mediafire.com/?zwcuj0laydh6cqp
|
|
This is nothing short of amazing. RET FIGHTING!
|
As some people pointed out the warpgate mechanic lowers that metric for protoss. You are supposed to have money one the warpgate cooldown is finished - if not it is even worse for your macro... I think it is also ok for zerg to have floating money - Due to their mechanics they don't lose larva and can spend vast amounts of money very quickly when needed...
|
Amazing, thank you so much!
|
On October 29 2011 16:09 Nightmer09 wrote: and ofc, eu>na
lmao, euro pride is fucking hilarious. But wow, demuslim coming out ahead of even KR players, thats pretty crazy. I always thought he was a great, but not top tier player. Maybe i was wrong and his results will match his skills soon
|
That was really interesting reading! Good job sir!
|
i keep getting floored with how much effort and energy everyone is putting into data collection here at TL.
Every time i think we've hit an end, another study comes out, and shows how much better x person plays.
I love it.
|
3 of the top 5 on the american ladderboard are accounts from european pros, already present on the europe ladder as well. Should they be counted in?
|
IdrA definitely is on his game at tournaments more than ladder.
And I'm glad to see DeMuslim way up there. I've been waiting and waiting to see him in some tournaments, because I know he's awesome. xD
|
HuK looks like Anton Chigurh (javier bardem) from "No Country for Old Men" in that pic
|
Tod is the best !
|
This is very intriguing. I'll be interested in seeing more results from the other regions. Nevertheless, great work!
I just wish I could remember to keep adding my SQ after every game so that I could try to compare my spending to the spending of pros. My SQ is all over the place though, with a range of like 60-88. This is definitely a good indicator of whether I'm improving my spending though! Thanks for all the time you've put into this.
|
people keep bringing up protoss SQ but what about zerg? Isn't it interesting that terran is always the highest? I think this may be because zerg maxes out quicker and ends up stacking a lot of minerals/gas and even after remaxing still has problems spending efficiently. Also muta builds eat into the SQ, because like warp gate you sometimes stack up to 1000/1000.
I also wonder how larva effects SQ for zerg. Zerg might have the money to spend but doesn't quite have the larva for it.
|
On October 29 2011 17:18 emc wrote: people keep bringing up protoss SQ but what about zerg? Isn't it interesting that terran is always the highest? I think this may be because zerg maxes out quicker and ends up stacking a lot of minerals/gas and even after remaxing still has problems spending efficiently. Also muta builds eat into the SQ, because like warp gate you sometimes stack up to 1000/1000.
I also wonder how larva effects SQ for zerg. Zerg might have the money to spend but doesn't quite have the larva for it.
zergs naturally will bank resources a bit more than terran because they tend on mass producing an army or drones on certain cues in the early-mid game. terran on the other hand is constant production of workers/army so it's always bad to bank resources. actually, the major part of terran macro is knowing when to build more production facilities (and constantly producing), while this is not really relevant to zergs.
this is pretty interesting. I just calculated sq for a bunch of my games, and the matchup that i'm very confident in actually has the highest sq (88~ average vs 78/76 averages)... and my total average falls in with other data too.
|
A tool that can scan a set of replays and compute the SQ would be nice.
|
Idra beasting it up when it actually matters. Highest avg SQ in MLG, SICK!
|
Nice analysis. I'm actually very worried that with BW teams possibly switching to SC2 and already practicing that the skill ceiling may be reached really soon, with players like Bisu, Jaedong, Flash, etc...
I mean I would love for blizzard to make the macro mechanics harder.
Mules could last half of what they last now, of course the energy cost cut in half as well, larva inject benefits cut in half, but of course the ability made every 20 seconds and I think protoss are good in that regard.
|
lol of course goody would be 5th on eu since he always ques SCV's and Tanks 3 times. I think its important for terran players to analyze their replays and also report how much they queue along their SQ.
|
very nice, can you now norm the sq so that max sq is... say 10000, so that we then can claim that bomber's macro is OVER 9000!!
|
Nice analysis, but have you tried to plot SQ by race, because their mechanics are different and naturally terrans should have lower unspent ressources, with terrans you should never have unspent ressources whereas with protoss and zerg it's inevitable. I don't if it would be legitimate to do so, but maybe you could try to find a coefficient for each race that would allow to compare the SQ between different races?
|
On October 29 2011 11:12 Theeakoz wrote: This is amazing, thanks. I love demuslime always have, who is faith? funny haven't heard much of him.
Faith is on team VP...although I guess now he is part of CheckSix because of the merger. He is really good. Of all the VP players I like Faith and Chance the best. VPChance is a beast zerg. If you have time you should check them both out. Great to learn from.
|
Man this is so cool. I had read the earlier one where you explain the Spending Quotient and that one was great but this really shows the potential it has, and how awesome the information is you get from it. On top of that I'm close to jealous of the "happy feeling" you must have from actually creating something like this that is legit and in my eyes has a lot of future perspective in terms of becoming an established metric that is recognized by the majority of the community.
|
Incredible analysis! I wonder if the lower protoss average SQ from orlando is because of the amount of pvp's, which have messed up spending. for half of the games, half of the time you're pooling money (four gates) and mc and huk played eachother more than the others did on average...
|
Would it be possible to post up the t-test calculations? Great write-up man.
|
Now we need some way to measure production buildings utilization.
|
WEll protoss have a button where they can macro. ITs called W.
|
Very interesting post. The information was laid out very nicely. Thanks for the hard work.
QUESTION: Is it possible AND would it be valuable to map out SQ for of individual players as a function of game duration?
I ask for the following reasons:
1. Most of these players have very refined builds as such SQ during the first 10 or so minutes is probably quite high.
2. If high SQ does not have strong correspondence to wins of short duration games then it might indicate that micro/harassment/timings play a stronger role in winning during the early game.
3. It is possible that in long duration games well-rehearsed builds give way to macro-skill as players adapt to the game situation and draw upon either good mechanics or game knowledge or both. If those skills which probably result in high SQ win more long duration games, then it might prove to be valuable information.
I recognize that SQ has several limitations, but additional analysis of SQ in games of various duration might yield interesting results.
|
wow interesting! thanks for crunching the numbers!
|
|
|
France12454 Posts
I don't understand how this can measure macro since if you perma queue up you will have a good score? XD. Look at Goody for example. And what about zergs who eco for tech switches or whatever? How is it counted in average unspent ressource? (I mean it's average unspent every 1 minute, or 30sec, etc)
|
Germany3367 Posts
Wow yery interesting article. Thank you very much man, can´t wait for the other regions!
|
Fascinating. I'm actually quite surprised at how much effort went into all this number crunching and mind numbing statistical calculations. A truly outstanding effort. I can't wait until the report is complete.
|
Protoss Units have a higher cost, compared to other races, so our SQ has to be lower. You are able to build a marine with 50 Minerals. But as Protoss , you have to stockpile another 50 Minerals to build a zealot. Therefore the average unspent resource will always be higher as protoss, no matter what.
You need a additional factor here.
|
"t (p = 0.02, two-tailed t-test)"
more posts need to use shit like this
fantastic.
|
Very good read and analysis. I am wondering if maybe there was a possibility to come up with a more comprehensive "macro score"? Something that factors in other things, not just spending. The problem here is probably that you need information that can only come from inside the replay, so the data gathering would me much more difficult.
As for the racial difference, you might need some race coefficient in your formula - but I guess only a truly gargantuan sample size could properly determine such a thing.
How did you collect the data? Maybe there is an easy way to get a lot of replays and automatically get the data?
|
Team science fighting!
Really impressive work.
|
Nice stats
|
GoODy in front of IdrA in a Statistic about macro abilitys. Who would have thought that
|
On October 29 2011 11:15 K3Nyy wrote: Protoss' SQ is lower because there's really nothing you can sink your money into in between cycles.
Terran's SQ should be higher because they can queue units. Goody being anywhere on a list of being good at spending is proof of this. A terran can just mass queue units without needing to be spending efficiently and they come across as having a good SQ.
|
Maybe you shouldnt put european players on the american bracket, theres a lot of double players there. Otherwise great read ~!
|
Is Faith fayth from bw? If he even plays starcraft anymore, that is...
|
I'm blown away. Whatthefat, you could be TL's own analytical headmaster. The Data Lord.
|
Imo Zerg's SQ is actually little bit DEFLATED but that's just my opinion. Larva mechanic keeps them locked and They can't just throw down extra production facilities haphazardly. That's an interesting find if that is true for the Protoss players, but it might just be attributed to the fact that the warp gate mechanic does hinder their spending. Either way great write up, I really liked this and I'm going to post this on the battle.net forums if you don't mind! (A link that is )
|
This method of macro comparison is broken.. and doesn't consider fundamental elements involved in the act of 'Macro'.. such as the race requirements of spending money.. 200/200 situations.
Interesting concept - and its almost there.. but the mathematical problem is not solved with this formula.
|
Very Nice read good job
|
Love these threads, read it all. Hope to see Nestea's SQ ^^, Maybe you can use his Blizzcon results?
|
amazing.. thank you for putting so much work in this
|
When I read this last night I knew it had to be some sort of amazing analysis that I couldn't fully under how good it was.
When I got on this morning and saw it on the front page I knew I was right. haha great write up!!!
|
Love to see the SQ being adopted as a "semi-official" measuring stick
|
|
This is the most hilarious objectification of one primary skill in SC2. I just got trolled hard for the 5 minutes I actually spent looking this over. Thx
Why not analyze something meaningful like amp redundancy or how many times a player drools during a tournament game. How about, how many TvT are won with absolutely no micro, or analyze effective use of ctrl groups based on categories of unit compositions, anything but this BSQ. Like today, did TLO maintain groups of 7 vikings while they were being torn out of the sky by zerg after he opened with very fast muta den?
User was temp banned for this post.
|
Really impressive by DeMusliM and Rets, soo high those few games O_O
SQ was off the roof
|
I really would like to see a comparison of the servers across all the leagues! Great write-up, though!
|
I love your application of stats and math to differentiate pros spending. These are such interesting results. Thanks so much for putting in the time to do this.
|
The amusing thing is that while it should be a totall broken method of measuring things that shouldn't be applicable generally and should have pretty different results depending on race, in theory, the fact is that what seems obvious on the outside (terran having easiest time, zerg and protoss having a harder time due to mechanics), it doesn't seem to actually matter, and it actually seems to provide a meaningful result when you use many games and average out the numbers.
It's one of those things which should never work, but does seem to in practice.
It might be interesting to compare the top 10 terrans to the top 10 protosses to the top 10 zergs and see how that impacts things, and maybe consider the implications of a scaling method to cover comparisons between races, since there should be differences in theory.
|
numbers, numbers... everywhere!
and i <3 it. I love the analysis and look forward to future regions and more info added to the findings.
|
wow, man. very impressive study. a superb combination of sound science and in-depth knowledge of the game and the pro-scene. i really think you should do this full-time. and i also think TL should crowd-source small donations to fund studies like this one! pm me to discuss this!
|
Great article. I would love to see the SQ between MUs. A player should get a higher SQ when his income rate is higher. Taking in consideration that players can macro well with every income they have, players should not be able to get very high SQ when playing ZvZ (their income just isnt high enough to get a high SQ). Would like to see if PvP has lower SQ because of low eco play than other MUs (this could also explain HuK and MC scoring low on MLG, they has mostly PvP games)
PS. Maybe I do the math wrong, but dont think so.
|
You gotta remember that goody's money is low cos he has 4 thors queued up per factory
|
Awesome article! Really thoroughly done.
|
On October 29 2011 20:15 alepov wrote: You gotta remember that goody's money is low cos he has 4 thors queued up per factory
lol. It's funny but it's actually not a bad point. Shame there isn't a way to access the replay files and include queued units in unspent resources.
|
Extremely good work! Well, I guess coaching should focus even more on "pylons & probes, pylons & probes" and "for every x money, you must build a macro hatch!" now
|
i dont really believe in this numbers but nice for the ego. in avg im better than most korean pros ^^
|
As I scientist, these type of articles warm my heart. Great writeup. Surprised that Idra's macro is THAT good. I thought it was overhyped. Hoping to see more SC2metrics soon.
|
First of all: great article, good effort, but wow im TOTALLY confused by the axes.
What does the left axis with the games mean? Please some give me an example on how to read a graph.
Isnt the games number just the sample size? That shouldn't be in the graph as one of the axes then. Im totally confused, please help lol :D
|
On October 29 2011 20:39 MasterReY wrote: First of all: great article, good effort, but wow im TOTALLY confused by the axes.
What does the left axis with the games mean? Please some give me an example on how to read a graph.
Isnt the games number just the sample size? That shouldn't be in the graph as one of the axes then. Im totally confused, please help lol :D the amount of games with the corresponding SQ. So, for example, ToD has submitted 15 games with 90 SQ
|
Wow, this is incredibly evenly distributed.
|
On October 29 2011 20:46 alepov wrote:Show nested quote +On October 29 2011 20:39 MasterReY wrote: First of all: great article, good effort, but wow im TOTALLY confused by the axes.
What does the left axis with the games mean? Please some give me an example on how to read a graph.
Isnt the games number just the sample size? That shouldn't be in the graph as one of the axes then. Im totally confused, please help lol :D the amount of games with the corresponding SQ. So, for example, ToD has submitted 15 games with 90 SQ
Yeah i thought of that too, but by that logic he also submitted 10 games with SQ 85 and 95 ??? Makes no sense t-t
|
On October 29 2011 20:52 MasterReY wrote:Show nested quote +On October 29 2011 20:46 alepov wrote:On October 29 2011 20:39 MasterReY wrote: First of all: great article, good effort, but wow im TOTALLY confused by the axes.
What does the left axis with the games mean? Please some give me an example on how to read a graph.
Isnt the games number just the sample size? That shouldn't be in the graph as one of the axes then. Im totally confused, please help lol :D the amount of games with the corresponding SQ. So, for example, ToD has submitted 15 games with 90 SQ Yeah i thought of that too, but by that logic he also submitted 10 games with SQ 85 and 95 ??? Makes no sense t-t No, you can see the points where the graph changes steepness, the rest is just a fit. So he also submitted ~6 games with ~82 and ~5 games with ~98. (on the EU graph)
|
Sexy, I agree with the majority of people when saying protoss's SQ is slightly higher because we cant warp in between cycles the top scoring games in terms of SQ both featured a large amount of collosi.
|
Nice work, great article! Well done sir.
|
On October 29 2011 11:11 Whitewing wrote: One issue I have:
Zergs, when going for muta builds, deliberately pool resources and float them for a while before the mutas come out, which raises their average unspent resources a fair bit. Going for infesters they float a lot of gas as well, etc.
Protoss play a heavy gateway style often will gain 600+ minerals and a bunch of gas too before their warp gates come off cooldown and then they burn them all at once. That would raise their average unspent resources.
Terran is the only race that doesn't float a lot of resources by design.
One should compare the metric between people playing the same race, but not compare across different races.
I concur.
|
Awesome write up. Gonna read it through when I got time!
|
Seeing GoOdy with such a high SQ reveals an obvious flaw in the system at the moment. To get a truly accurate gauge you would need to count resources tied up in queued units as unspent, but that would mean that require so much work and is not really feasible unless Blizzard implement this, or there is a tool that can analyse replays for this information.
edit: I also thought it was interesting how IdrA really seems to turn it on for tournaments. I was disappointed with his score on the ladder, considering that his reputation is as one of the best macro Zergs in the world, but seeing his score at MLG, particularly against Bomber you can see that he certainly does deserve his reputation.
|
United Kingdom14464 Posts
Awesome article, that was really impressive. it also confirms what I knew before: Demuslim is a damn macro monster!
|
On October 29 2011 20:57 alepov wrote:Show nested quote +On October 29 2011 20:52 MasterReY wrote:On October 29 2011 20:46 alepov wrote:On October 29 2011 20:39 MasterReY wrote: First of all: great article, good effort, but wow im TOTALLY confused by the axes.
What does the left axis with the games mean? Please some give me an example on how to read a graph.
Isnt the games number just the sample size? That shouldn't be in the graph as one of the axes then. Im totally confused, please help lol :D the amount of games with the corresponding SQ. So, for example, ToD has submitted 15 games with 90 SQ Yeah i thought of that too, but by that logic he also submitted 10 games with SQ 85 and 95 ??? Makes no sense t-t No, you can see the points where the graph changes steepness, the rest is just a fit. So he also submitted ~6 games with ~82 and ~5 games with ~98. (on the EU graph)
So ToD submitted 15 games with SQ of 90, 6 games with SQ of 82, 5 games with SQ of 98 and 1 game with SQ of 106. So he submitted 27 games overall (for EU graph)? I guess it makes some sense, but it just feels kind of unnecessary.
Anyways thanks^^
|
These kinds of measurements would be a powerful tool for a betting platform. They could be used to compute the betting coefficient for pro tournament games. Dunno how TL bets works, but TLPD is just useless for such things.
Of course there are numerous other factors that cant be addressed through a research like prayers mental status, preparation, nervs, etc..
Cheerz
|
I don't understand people arguing that protoss' SQ are lower because of the race mecanics, yet ToD has one of the best SQ of all time !
|
Ret and Bomber so gosu. I absolutely love these statistics man, really well done. I am going to have to calculate mine now, as I am convinced I am the best player in the world without knowing it truly yet(!)
|
Goody is up there because he queues 5 siege tanks in each of his factories. Amazing article by the way
|
on the subject of lower SQ for protoss. ive always felt that protoss players just flat out dont play as well as other races. the lack of multipronged attacks (omg a protoss made 1 warp prism = best player ever) many over commit on pushes etc etc so for me this sq seems to support my idea that protoss players have been carried by race far too long : D.
but on a more fair side of things, its probably worth noting that even the best macro terrans queue up new units ~10% before the first one is finished, on gateways this just isnt possible. so unless im missing something and sq factors out queued units, this ~15% fluctuation in sq between protoss and terran players could be purely down to more unit queuing by terrans.
zerg for all their 'hardships' actually has a reletively easy time spending their money, with the hatchery production system you can burn through 2k unspent gas in 10 seconds, im not saying it would be effective play, merely that it would raise your sq
|
Pretty interesting. But I have 2 questions:
1. More cheese on NA Ladder would would mean short games with low eco/money easy to keep low. 2. Wider range of good/macro players on EU would mean more 20+ minute games where money piles up.
Few games wouldnt matter but the playstyle is different.
|
On October 29 2011 21:27 WhiteDog wrote: I don't understand people arguing that protoss' SQ are lower because of the race mecanics, yet ToD has one of the best SQ of all time !
The logic is that protoss warp ins have a long cooldown, so you basically warp in - let your many go up in the thousands on 6+ gates, and then warp in and go down to 0. Over and over. This basic mechanic makes protoss look worse at macro in these things. Also our units are the most expensive, making us pool more money before we spend it. Compare it to someone making mass marine, who can - in theory - keep his minerals constantly below 50 minerals.
Edit. I didn't find it in the op, but are games also evaluated that go to 200/200? Because then spending dramatically decreases in comparison to income rate. Didn't Nestea have like 14k banked?
|
I think the protoss deficit is largely contributed them having idle gateways deciding weather or not to warp in the next round of units, or continue building more tech/upgrades/robo units.
This is clearly noticeable especially in Huk's PvZ where he'll power up to a ton of gates when he gets his third up.
|
The fact that protoss score lower in sq than other races, can this be because it is harder to reach an incredibly high income? Terran missing mules? Drop them all at once, if your orbital energy didn't exceed 200 and continuously gained energy, you have the same income on average! Protoss missing chronoboosts on the nexus? Guess your macro will be falling behind. (Zerg also drone like crazy, that's how the race is suppose to be played, no explanation needed.)
|
This is incredible, I can't wait until you add the other regions.
|
On October 29 2011 22:03 CatNzHat wrote: I think the protoss deficit is largely contributed them having idle gateways deciding weather or not to warp in the next round of units, or continue building more tech/upgrades/robo units.
This is clearly noticeable especially in Huk's PvZ where he'll power up to a ton of gates when he gets his third up.
Not really. The SQ doesn't care what the money is spent on, as long as it's spent. Overmaking production structures would, if anything, tend to increase the SQ artificially (in relation to game skill, since it's a mistake) because macro mistakes and missed production cycles would be more easily forgiven.
People have already noted that Protoss tends to float resources between production cycles, just due to the nature of the warpgate mechanic, and then spend it all at once, whereas, with other races, it's easier to just build a single unit immediately you have the resources for it. That's the most likely cause of a slight deficit here.
|
On October 29 2011 21:53 Euronyme wrote:Show nested quote +On October 29 2011 21:27 WhiteDog wrote: I don't understand people arguing that protoss' SQ are lower because of the race mecanics, yet ToD has one of the best SQ of all time ! The logic is that protoss warp ins have a long cooldown, so you basically warp in - let your many go up in the thousands on 6+ gates, and then warp in and go down to 0. Over and over. This basic mechanic makes protoss look worse at macro in these things. Also our units are the most expensive, making us pool more money before we spend it. Compare it to someone making mass marine, who can - in theory - keep his minerals constantly below 50 minerals. Edit. I didn't find it in the op, but are games also evaluated that go to 200/200? Because then spending dramatically decreases in comparison to income rate. Didn't Nestea have like 14k banked?
If you had 5 games that all go 200/200 and they turtle up for 10 minutes then it would skew the data of course but the data is collected from A LOT of replays that average out with players spending and trading armies accordingly. At the top level it's rare to see players sit at 200/200 anyway so I don't think that's a problem.
|
protoss warp-cycle mechanics dont allow for a similar SQ since you cant queue up shit. noone has perfect warp-in cycles, but people have a higher SQ for queueing up shit. also you need to change your view to the next pylon to warp in, while as a terran or a zerg you can reinforce by just hitting the hotkeys while having view of the fight, that's why you generally float a little bit more money infight as a protoss.
just check out goody with his embarassing macro, having one of the highest SQs because he queues up 4-5 tanks in each of his factories. i think the SQ is quite irrelevant in terms of comparing macro between races, but statistical information is always great to have!
thanks for the effort, interesting read.
|
wow very cool nice Saturday morning read I love stuff like this keep it up!
|
Bitbybit.Prime would have the lowest highest SQ in the world. By far.
|
Can't wait for the Korean SQs.
I expect Protoss to be lower there, too because of how they stack minerals before each cycle
|
Interesting. Can't wait to see rest of the regions. Good job and congrats for your TL writer spot!
|
Very nice analyse but i see a problem in the unspent ressources.
If you are maxed for a couple of minutes (and you save your ressources) the unspent money explodes. Then it is not a viable methode anymore how good you macro is.
|
On October 29 2011 22:38 Euronyme wrote: Bitbybit.Prime would have the lowest highest SQ in the world. By far.
Oh really?
|
Very impressive! Keep it coming!
|
I'd like to see some analysis on the argument that SQ is less useful or even sometimes not beneficial to protoss, owing to their gas dependancy, the tendency to stockpile unspent minerals is often unavoidable or indeed beneficial (allows the instant production of the required amount of wall+cannons for an expansion or the huge reinforcement warpins, although this factor also applies to zerg play.) It'd be nice to see if you could find any mathematical way of analysing this, as it seems your data on the protoss race shows a discrepancy that goes far beyond variance.
|
Sadly, SQ isnt anything at all. I have an avg sq of 78 and im gold. My max was 94, and about 30% of my games are in grandmaster level SQs. The problem is, I dont get the right units and i get them too slow, eventhough in a lot of games my income is almost equivalent to a masters player.
Another thing is that i loose a ton of games vs harassment, dts, mutas etc. Thats what really is important in SC2, the SQ is almost nothing; I wish it were different
In top notch pro levels, the SQ doesnt tell alot either. Obviously you cant win mlg w/ a spending quotient of 30, but it does not matter if you have 90 or 100, as the scale is nonlinear, especially for high SQs and Avg Incomes.
|
Amazing read, but goody on a list for GOOD macro? something is not right here..
|
i'm glad this made it it on the front page of TL. i read the original post too, and this one just stunned me again :D
|
if you want to warp in 5 stalkers, you need to pile up 625/250, for every warp gate cycle you need another 625/250, all game long.
you can't use SQ to compare players from diferent races. that is obvious.
|
On October 29 2011 22:59 nucLeaRTV wrote:Show nested quote +On October 29 2011 22:38 Euronyme wrote: Bitbybit.Prime would have the lowest highest SQ in the world. By far. Oh really?
Well obviously? He doesn't have any minerals, so it'll be through the roof!
|
Can I ask how do you calculate it ? I don't have much time atm, so I took a quick look at the thread and couldn't find it ^^ From what I saw, great post !
|
|
Great work man!
Btw, in AM server, EGDeMusLiM is actually EGPuma (or shared account with Puma from what I heard), and BcuzofuProS is FnaticRain. Shared account between progamers might be the reason some players have different SQ between two servers or between ladder and tournament.
|
How can SQbe accurate for Zerg? Zerg is based around pooling minerals in the mid game to make 15+ units at a time. And In late game you must pool enough minerals to insta remax.
I think if idra was Terran or toss his sq would be highest.
|
Bomber has the highest SQ....of course. and Idra is pretty far up there too, hes a macro/mechanics beast
|
On October 29 2011 23:46 Cosmology wrote: How can SQbe accurate for Zerg? Zerg is based around pooling minerals in the mid game to make 15+ units at a time. And In late game you must pool enough minerals to insta remax.
I think if idra was Terran or toss his sq would be highest. How would IdrA's SQ be the highest as Protoss? You need to let resources accumulate until you warp in a whole production cycle. Did you miss the whole discussion in this thread of why Protoss players have lower SQs?
|
On October 29 2011 23:07 CBenni wrote:Sadly, SQ isnt anything at all. I have an avg sq of 78 and im gold. My max was 94, and about 30% of my games are in grandmaster level SQs. The problem is, I dont get the right units and i get them too slow, eventhough in a lot of games my income is almost equivalent to a masters player. Another thing is that i loose a ton of games vs harassment, dts, mutas etc. Thats what really is important in SC2, the SQ is almost nothing; I wish it were different In top notch pro levels, the SQ doesnt tell alot either. Obviously you cant win mlg w/ a spending quotient of 30, but it does not matter if you have 90 or 100, as the scale is nonlinear, especially for high SQs and Avg Incomes.
This likely has to do with a lower overall income, not poor unit selection. It's still poor macro. It's funny, since I first read this guys theory I've noticed it a lot in the score screens of my 4v4r games. The higher the league the player is, you'll see a higher income generated and lower floated mineral total at the end. And it's rather significant. The gap between gold -> masters (the range of players i play) is almost entirely macro oriented, not micro like most people at the gold level tend to think. Sure, micro comes into play at times, but in general terms, the better macro players get higher rankings in the end.
|
Very impressive analysis. The next frontier seems to be distinguishing between deliberate pooling and macro slips.
|
I think the reason you don't see a huge difference between P and T is because even though it is possible for T to queue up as soon as they have the resources too do so, they don't. Players queue in waves, as units are almost complete. If you are producing marines out of 3 rax you will queue all the rax at the same time, not each marine as you get 50 minerals.
|
The protoss bias is quite clear even without the other regional data, but would be nice to confirm it.
The different production cicles for each race mean we should analyze more data to find trends, and in the end it should result in finding a factor to weigh the results depending on each race. I for example find that the way terran can queue up units in their production buildings can be also an underlying factor to a higher SQ.
What to do? analyze more data, find trends, and then find a weighing factor for each race. I like this article a lot! +1
|
On October 30 2011 00:22 hypercube wrote: Very impressive analysis. The next frontier seems to be distinguishing between deliberate pooling and macro slips.
That should be put into thought, and simply remove from the sample every game that goes above a certain parameters like, money piling when some player is 200/200
|
Great analysis man very impressed . Just a couple of points that maybe dont show up in the analysis:
1.SQ probably scewed in Terrans favour by their ability to queue units 2.SQ might disfavour the other two races slightly more, due to the 'rounds of units' mechanic used by Protoss and Zerg (Warpgate rounds and production rounds after lava injects)
Hard to say how to account for these problems though
Overall great stuff, nice to see how in-depth we can start to quantify a players ability to macro. Also good to see my fellow Brit DeMuslim topping the charts
|
Really amazing stuff, well obviously SQ isn't the end all be all, it is still a cool thing to look at. I think it would be really amazing if the research could be expanded to include not only the other ladders but other major tournaments including past mlgs, gsl ect.
|
so basically ur saying Ret is good?
Thanks, good to know :D
j/k this is an amazing piece of work and its clear uve spent some time and brainpower on it, i dont believe any of ur findings really prove anything or make any sort of ground breaking discovery in terms of how skill works and winning in sc2 works or anything like that, but it is deffinitely interesting information to play around with
|
Wow, thats such an interesting take on skill level, I remember hearing Huk say that his best game was against STJuly too, strangely enough that was his SQ peak, should be a good gauge on when players are slumping as kind of an average of what their SQ was at and what it's at now...
|
Great write-up. Love the "Top 3 Control, Bitches!" haha
|
Sick research, amazing work! Can't wait to see all the stats when everything's done.
|
Goody a top5 macro player in EU? I actually enjoy Goody (one of the few probably) but hes not a good macro player. not close. I'm not convinced by SQ being a relevant index of macro skills really. Part of it but somethings missing
|
If I queue up my production for five units will that give me a higher SQ than a player who never queues up units?
|
|
thanks!
can't wait to see the differences across the azn servers
|
Does the scale account for the Terran's ability to queue units? You can put up to 5 units in the building's queue, which counts as 'spent' but that would contaminate your end results. I loved reading the article, it was like a scientific paper.
|
France12454 Posts
On October 30 2011 00:30 sirrobert5 wrote:Great analysis man very impressed . Just a couple of points that maybe dont show up in the analysis: 1.SQ probably scewed in Terrans favour by their ability to queue units 2.SQ might disfavour the other two races slightly more, due to the 'rounds of units' mechanic used by Protoss and Zerg (Warpgate rounds and production rounds after lava injects) Hard to say how to account for these problems though Overall great stuff, nice to see how in-depth we can start to quantify a players ability to macro. Also good to see my fellow Brit DeMuslim topping the charts But terran income is higher so it's hard to spent more ressources? XD. So yeah basically these data are not very meaningful
|
SQ is not a particularely great indicator imo but it does seem to have a decent correleration with skill level. It also seems to portray some decent features that it's relatively independant of average game time and race making it easy to measure. Still all in all it seems a bit of a useless statistic.
APM however is also quite the useless statistic and still get's cited a lot so there is still a good chance this might linger around.
|
i love how it shows goody as a solid macro player. money in the cue is money spent, right? still, amazing read. thanks
|
I would like to see how SQ correlate with game time! please make it happen dear OP.
|
absolutely awesome write up! took me forever to read it all, but it was totally worth the time! keep up the good work!
|
This was really cool.
Everyone's talking about how Protoss pool money with warpgate research. I think it would be interesting to see if a Protoss who didn't get warpgate and just queued units like a Terran had a higher SQ. Could be interesting and it would prove whether the warpgate mechanics results in a lower SQ.
|
On October 30 2011 00:54 IzuN wrote: Goody a top5 macro player in EU? I actually enjoy Goody (one of the few probably) but hes not a good macro player. not close. I'm not convinced by SQ being a relevant index of macro skills really. Part of it but somethings missing
He queues a lot = Low unspent resources.
|
Nice analysis!
I am surprised, as many other here, that you don't see a larger SQ for terran at all levels, due to queuing. Weird! Does that finally prove, as we all suspect deep inside, that terrans are noobs with inflated MMR due to imbalance?
Looking forward to see if the koreans mechanics are indeed at a higher level.
And it'd be nice to see nesteas SQ in that blizzcon game where he stacked over 10k resources a large fraction of the game. Bronze level SQ in blizzcon final?
|
Goody is 5 in EU? I guess 12 queued siege tanks explains that?
|
Wow well done. Great read, very interesting!
|
Just finished the read, was really interesting and loved it. Thanks for putting in the time, it looks like you really worked hard on this. Also we learned that the key to spending your money is breaking your arm's as much as possible, I see now lol... jkjk. Great Job!!!! : )
|
Nicely done !
|
Really nice article. I congratulate you on your effort. And it's great to see how do top players spend their resources.
|
This was probably said a million times already but if not: Protoss can't queue gateways. It would make sense that they have a worse SQ since their main unit producing structures can't queue, and can only make one unit at a time.
|
I wonder if you could pull stats from thousands of mid-level players for Protoss/Terran/Zerg and readjust based on Protoss having more minerals as a standard.
|
On October 30 2011 02:38 esaul17 wrote: This was probably said a million times already but if not: Protoss can't queue gateways. It would make sense that they have a worse SQ since their main unit producing structures can't queue, and can only make one unit at a time.
gateways are 1 of 4 producing structures what about stargates, nexus and robotics. carriers can also que..
|
That's a lot of work...!! Thx! Nice read
|
Nice Read, i looked over the original SQ post too. Statistics application into sc2!!
|
On October 30 2011 02:44 Liquid`Nazgul wrote: I wonder if you could pull stats from thousands of mid-level players for Protoss/Terran/Zerg and readjust based on Protoss having more minerals as a standard.
I don't see why you would though? I mean, assuming you don't queue as a Terran, why would a protoss have more minerals?
|
from a math fanatic to another this was a very interesting read thank you. im off to look at my SQ now :D
|
Edit: accidental post. how do you delete posts?
|
Holy shit! Idra, Ret, Bomber and the STC are so crazy efficient. Bomber especially is just crazy good.
|
i like this, it really shows some things.
however you can't compare this between races, as for example terrans are the only race that should NEVER get too many minerals stacked up unless they get maxed, while protoss and especially zerg sometimes bank minerals. there may be exceptions though sometimes!
still very cool
|
Nice to see TL embracing WTF.
Overall, another fantastic article by WTF.
|
most of protoss spending goes into warp in rounds so yeah. but i guess this can be used to compare players of the same race.
|
This is pretty nice.
I did some testing in single player vs very easy AI macroing my brains out with 10 bases and bazillion overlords managing a 137 SQ and I have to say that if it's correct, that 121 SQ by Ret is absolutely incredible. :O
|
Do you take into account the average cost of units per race into the calculation? If Zerg and Terran both spend 50 minerals per tier 1 unit, a Protoss player would have to wait twice as long to build a tier one unit.
|
Statistics ! omnomnomnom
The difference in highest SQ in a game and average SQ is pretty high. I wonder if SQ could be a good way to measure the player's commitment to the game and shape of the day/moment. Comparing SQ and APM on a game to game basis should reveal some info on that.
|
there are several situation where Z saves up ressources and larvae (waiting for spire to finish or when it is unclear wether a timing push is incoming e.g. 4 gate) so Z sq might be somewhat lower in average.
|
Fantastic data spread and analysis. And well written article.
The only thing I really have to say though is, well, DUH. Just like the last article, all this does is show that "better players are better at the game" - be it unspent resources, making workers, killing units, killing buildings, winning games... the top 10 players are top 10 for a reason. I am not sure if anything different was even expected? It's cute, just not interesting... yet!
The next steps are obviously to compare between races and within races, identify statistics that correspond to certain play styles, possibly even analyze these stats at different points during a game and also if they depend on game length (if this is possible), and then start developing relationships this way. Then some real knowledge would begin to emerge regarding the metagame, not just "better players are better."
edit: and the huk/protoss low SQ thing is the right direction to go in. this is where the knowledge is!
|
Once the data set gets larger, it needs to be broken down by the 6 matchups. I get a feeling some match ups will favor skills that aren't captured in a straight macro mechanic.
Oh, and Ret having a 121 was zero surprise, haha.
|
Really nice article. Ret still macro god, obv. Nice to see Haypro on the list. At least he can macro.
|
Nice stats! Definitely interested in seeing what it is like at the other regions and more comparison of players who play in multiple regions (like what you have done w Demuslim and Tod)!
|
great article protoss SQ is probably low because you warp in.... oh well. Huk ftw!
|
Haha, the "GoOdy Q" prevails. I guess having 5 tanks in queue at one factory fucks up your SQ results.
|
People are mentioning the warp gate mechanic, but it might have to do with the cost of the units as well. With the units on average more expensive for Protoss, the unspent resources would tend to float higher when buying them wouldn't it? Might be negligible though.
The other thing I've seen people doing while playing Protoss is that with the warpgate mechanic rather than continuously producing units they will stock pile warpgates and wait for battles to warp in whole defensive armies. This would also lead to a greater amount of floating resources.
|
On October 29 2011 22:13 Aim Here wrote:Show nested quote +On October 29 2011 22:03 CatNzHat wrote: I think the protoss deficit is largely contributed them having idle gateways deciding weather or not to warp in the next round of units, or continue building more tech/upgrades/robo units.
This is clearly noticeable especially in Huk's PvZ where he'll power up to a ton of gates when he gets his third up. Not really. The SQ doesn't care what the money is spent on, as long as it's spent. Overmaking production structures would, if anything, tend to increase the SQ artificially (in relation to game skill, since it's a mistake) because macro mistakes and missed production cycles would be more easily forgiven. People have already noted that Protoss tends to float resources between production cycles, just due to the nature of the warpgate mechanic, and then spend it all at once, whereas, with other races, it's easier to just build a single unit immediately you have the resources for it. That's the most likely cause of a slight deficit here.
have you ever played zerg? u always inject in cycles which means u will also float resources as zerg between production cycles.
It's only terran this doesn't affect because they can stuck up units as fast they get more minerals. I'd say Ret and idrA is up there only cos they are macro beasts and have superior macro than the others.
|
Wow really well done, did you consider doing a fit to the exponential data? You could determine how much players macro increases/decreases with game length for each demographic. Also, I think that makes sense how protoss is lower just due to having to wait for warpgate cooldown which inflates their average unspent for the time it is on cooldown before they dump their resources.
Also something that might be interesting if you got a large enough data set for a few players, you could use a Fourier transform (FFT command on MATLAB) and determine what the frequency modes of their macro is, e.g. the numerical difference for how they do in tourney's vs ladder. Really cool study.
|
Too bad you can't auto-analyze replays to not account for unspent income if you ever are at the foodcap or so.
|
Excellent write up thanks very much, i was always wondering about a method of measuring macro and this seems pretty damn good. also awesome huk pic at the end :D.
|
Very interesting article. It's nice to see how players at the top are absolutely great at spending compared to players below them without even accounting for the fact that lower level players may not have optimal income in the first place. Pretty eye-opening.
I would love to see top level averages for each race in a similar fashion to what you did with each player as a comparison between the races (not to say "X race players all suck", but to merely observe what levels of SQ are considered top-level for each race; certainly, it would be different for each race because of the vastly different mechanics).
|
You should talk to the guys who make Sc2 Gears. They should be able to implement this pretty easy, Sc2 gears already pulls all the info, it just needs to be run through the formula.
|
I want to add a plea for people to come forward and do the data collection for the other servers. I can't wait to see how the Koreans will do!
Unfortunately, the only way to collect the data (that we know of at the moment) is to use the GM ladder in Battlenet and look at the match history for players, and note down the data in Excel. It takes a couple of hours, but it's worth it.
The MLG tournament analysis is really interesting and it's nice to see that players seem to up their skill for a big event.
I'm glad that people are checking out their own SQs. I've been using this to inspire me and since I've been noting down my SQ after each game my average has gone from 45 to 65.
Some important things I want to note about game length though: 1.If the average income is less than 600, this game is disarded as the results aren't reliable enough for calculation SQ (see whatthefat's previous thread). 2. If your income is over 2000, then the calculation of SQ is very forgiving - you can have an upspent minerals of over 2000 and still have a good SQ score. This way, even if you're maxed out, it's not going to suddenly drop your SQ score unless you sit there for several minutes with 200 supply doing nothing. The pro's would be attacking and rebuilding their army at this point, so it's not going to make SQ less meaningful just because you hit max.
In short - the mechanic used for recording SQ already takes into account the differences in very short and very long games.
Also, for people wanting to measure their macro skills, the other easily measurable variable is number of workers (and number of workers per minute). At lower levels, you may be spending your money, but you will still struggle if your number of workers is low.
On October 29 2011 15:25 d9mmdi wrote: as you found out that your SQ is only one side of the coin (see Huk) maybe you could device a similar measure for control by using the units lost / resources lost data. Would probably require similar effort to normalize but would be interesting if you can analyze players in different categories. Might even be used in the predictor project to increase accuracy.
I think this is a great idea to analyse effiency - e.g. units killed versus units lost, but the data in the game summary is "units trained" and "killed units" and I think it would require "resources lost" to make any useful comparison. Zerg would lose out horribly because they lose so many zerglings.
|
It's pretty astounding that Ret and Idra can even be close to the terrans. Obvously protoss has no chance with how warp in forces you to pool money but even zerg with 4+ queens will pool money for every inject round. Not to mention that an important pat fo zerg is saving up a certain ammount of larvae midgame until you scout or deem it safe to use that for drones or your preffered unit. Obviously in short games zerg can be very effective but when a lot of zerg lategame strategy is based on pooling 5k+ mins and 2k+ gas to insta remax then obviously zergs are going to lose some rating points too. Actually zerg probably pools even more money than protoss late game. The only difference is that protoss pools money during the entire game regardless of income so obviously their unspent resources are higher throughtout the game.
|
Wow this is so amazing. For sure deserves the front page news and not just spotlight!
+ Show Spoiler +Top 3 control bitches : )
|
Wow, great article. Thanks for putting in the time to do all this work. I love the attention to statistics. Plus, it's fun to track my SQ over time and see how it compares to the top players.
|
I really hope you get school credit for this, if you did it for a class. It's just as informative as a few scientific journal articles I've read. Good stuff!
|
Maybe in the future you can find a way to convert sq between races. Seems like Protoss sq is slightly slower because of warpgates.
|
The joke about Demuslim's arm had me laughing IRL for like 5 minutes.
Amazing analysis, great work!
|
This is some great stuff, and I hope to see more of it in the future. I do think the SQ is affected by race mechanics, and hopefully that'll show up through more analysis. Warp-in and MULEs seem to have divergent effects on the SQ, as the former lends towards floating and the latter creates a fluctuating redline on SQ.
|
|
Nice job, proper scientific write up.
You know there's something there when the author can point out flaws in the theory, and it's still a strong idea.
|
Wow, this was freaking amazing stuff. Can't wait for the rest of the data from the other servers to really get an idea of how the different servers match-up. No surprise to see Bomber, Idra, and Ret put up their scores. Those kids macro like beast mode x100. But This maybe shows also how good DeMuslim really is and I can't wait to see him get even better in Korea! Great job! Very impressive statistics that we wouldn't normally be able to get in this amazing way!
|
United States10327 Posts
nice analysis
|
Wouldn't Terran's SQ naturally be higher simply because they can queue units up in their production facilities?
I'm not trying to take away from this excellent analysis by the way .
|
I really enjoyed this analysis and I look forward to reading about the other regions!
|
|
very interesting and well written article, thank you!
|
On October 30 2011 12:32 JeffVader wrote:Wouldn't Terran's SQ naturally be higher simply because they can queue units up in their production facilities? I'm not trying to take away from this excellent analysis by the way .
If you're aiming to purely increase your SQ then yes; however, since everyone in Ladder games (at least at the top level) isn't queing things up or spending just for the sake of spending, it shouldn't be a factor.
|
|
On October 30 2011 16:26 insufficient wrote: Tell that goody...?
Outliers happen; take the SQ as you take the BMI in regards to health; a good blanket overview of skill, however, not precise nor accurate in all cases.
|
Truly an amazing write up. Your attempt to understand the game better through the use of statistics (which aren't yet widely used outside of win rates) is both amazing and commendable.
There have been several comments regarding the Protoss need to float minerals for large warp-ins, as well as the Zerg's need to float fund in order to produce large numbers of mutas or infestors at once, which would of course affect SQ. As you said several times, SQ is not the be all-end all, but I agree with others in that it may be more telling if the players analyzed were from the same race, thus eliminating any fluctuations in the data due to different race mechanics.
Other than that, I love what you're doing, and you have nothing but my utmost respect for this study, and for all your work. Can't wait to see more data from different regions!!
|
United States32432 Posts
On October 30 2011 15:59 tgun wrote:Show nested quote +On October 30 2011 12:32 JeffVader wrote:Wouldn't Terran's SQ naturally be higher simply because they can queue units up in their production facilities? I'm not trying to take away from this excellent analysis by the way . If you're aiming to purely increase your SQ then yes; however, since everyone in Ladder games (at least at the top level) isn't queing things up or spending just for the sake of spending, it shouldn't be a factor.
Hopefully, one of whatthefat's future projects will be looking at SQ by race. I don't think it will tell us definitively what races are 'better' for spending money, but it might give us some insights into spending patterns.
|
its definitely biased against protoss and probably zerg as well.
protoss has warp gates which when macroed the resources have to accumulate up in cycles for them to warp in. warp gates also have no queue. like 6 gate would generally be banking up 750 resources pretty normally even with perfect macro.
zerg cannot queue up any units but larva do spawn constantly although there is also the inject cycles which at perfect macro encourage banking minerals again for those cycles.
terran can just queue up in all buildings which means that they can always keep money low even when macroing poorly and macroing perfectly still has no penalty for queueing up when the production numbers meet the income levels so money would theoretically stay much lower than other two races.
|
On October 30 2011 18:33 StrifeCro wrote: its definitely biased against protoss and probably zerg as well.
protoss has warp gates which when macroed the resources have to accumulate up in cycles for them to warp in. warp gates also have no queue. like 6 gate would generally be banking up 750 resources pretty normally even with perfect macro.
zerg cannot queue up any units but larva do spawn constantly although there is also the inject cycles which at perfect macro encourage banking minerals again for those cycles.
terran can just queue up in all buildings which means that they can always keep money low even when macroing poorly and macroing perfectly still has no penalty for queueing up when the production numbers meet the income levels so money would theoretically stay much lower than other two races.
Ideally, something like this would be done within constraints, say, only analysing a races SQ against each-other, and only taking into account games over a specific length, as it's much easier to keep a high SQ in short (<8:00) games.
|
Idra average SQ in MLG... Damn, its true as they say: His macro is insane, you dont want to go into a long game with him! (If you're not terran on shakuras plateau hint hint )
|
TOP 3 CON"TROLL" BITCHES!!! made my day
|
Wow, fantastic write up, well done!
I always though DeMuslim had impeccable macro when I watched him stream, low money all the time no matter what. Now its been proven.
|
Watching people summing 3 numbers and dividing by 3 to get an average and say: "This is statistics" makes me want to vomit. This article however is NOT one of those. Extremely nice article, which makes me happy that I have taken statistics courses so I actually understand what a 2-tailed t-test is :D Even though this is pretty "simple" statistics it is really well written and it does make a valid argument that SQ is a viable metric for assesing skill-level in general. Also: "Top 3 control bitches" made me crack up
|
On October 30 2011 18:33 StrifeCro wrote:+ Show Spoiler +its definitely biased against protoss and probably zerg as well.
protoss has warp gates which when macroed the resources have to accumulate up in cycles for them to warp in. warp gates also have no queue. like 6 gate would generally be banking up 750 resources pretty normally even with perfect macro.
zerg cannot queue up any units but larva do spawn constantly although there is also the inject cycles which at perfect macro encourage banking minerals again for those cycles.
terran can just queue up in all buildings which means that they can always keep money low even when macroing poorly and macroing perfectly still has no penalty for queueing up when the production numbers meet the income levels so money would theoretically stay much lower than other two races.
I was thinking this exactly, which is why SQ is probably only viable in the top of the ladder and at tournaments where doing that (pulling a GoOdy and queing up 5 thors) isn't gonna get you far, the "racial" bias also shows why HuK and MC could win even though they have lower SQ's. Because a perfectly macroing toss need to float minerals in cycles. Which is why you ideally can't asses the skills of different races using the same metric because the mechanics are way different.
|
nice analyzis, and as many people say the warpgate makes it biased against toss.
I like what Pokebunny said on the first page, that protoss is the only race where its good not to build units out of your gateways sometime.
But anyways, OP, you are a genius
|
I was about to ask about the possibility for an analysis between leagues but I saw your other post and had a read of that. Very nice and interesting statistics I will definitely be using this to work out my SQ in the future when I finish uni and finally have time to play enough games.
|
Doesnt such small sample size defeat the purplse of this analisys though?
I mean you can get different sets of last 30 games from the same player that provide wildly different results all day.
So what is this data for exactly?
|
On October 30 2011 20:21 ikona wrote: Doesnt such small sample size defeat the purplse of this analisys though?
I mean you can get different sets of last 30 games from the same player that provide wildly different results all day.
So what is this data for exactly?
For entertainment. Why so negative?
I really enjoy the idea of your project and it's an entertaining read overall.
|
Impressive article. A huge thanks to you, your study is pretty interesting, and the layout of your article makes it very enjoyable to read.
I can't wait for the Korean Ladder analysis, keep it up !
|
Oh I forgot to mention I am very keen for the other regions, any chance there would be a comparison with the leagues other than GM? Because as a player who plays on both SEA and NA I find it interesting that SEA seems harder around the platinum - diamond level while it is quite clear that NA has better players at the top level. The reasons for this aren't difficult to figure out, but it would be interesting to find out how the lower levels stack up.
|
On October 29 2011 10:54 whatthefat wrote:Show nested quote +On October 29 2011 10:50 MonkSEA wrote: I think the warpgate mechanic might actually be hindering Protoss' SQ, but very nice read. I'm pretty surprised at Bombers SQ and Ret's as well. That is what I suspect. Either that or something to do with the types of builds, e.g., it is beneficial to build multiple gateways all at once to sync up with warpgate research, meaning you have to temporarily float money. But I can't conclude that without more data.
As day9 pointed out on several occasions, if you start to float money which you need for your next warpin cycle, you should not instead buy structures you cannot really afford yet, since it will result in worse macro. Otherwise, by the time your warpin cycle comes you dont have enough money for units, and more structures than you can support.
This is a common macro error for protoss, and it will boost the SQ. Average unspent money will be lower if you instantly buy another gateway as soon as you have the money, and sq is not measuring your supply whatsoever.
But while this might be an explanation why lower level players can get relatively high SQ sometimes I doubt that this will be an explanation for differences at the pro level, where all the builds are usually more refined..
I'd rather suspect the latter possibility you mentioned, that for protoss, consistent builds tend to sacrifice spending efficiency for some reason.
|
Not sure anybody has pointed it out, but what happens if you take the PVP games out of it?
PVP is exclusively 1-base. Your spending quotient will therefore suck. In PvZ and PvT games, is pros' spending quotient comparable?
-Cross
|
On October 30 2011 22:30 Crosswind wrote: Not sure anybody has pointed it out, but what happens if you take the PVP games out of it?
PVP is exclusively 1-base. Your spending quotient will therefore suck. In PvZ and PvT games, is pros' spending quotient comparable?
-Cross
SQ would actually be higher in that case because the games are short and on 1-base, making it "easier" to spend all your money (more likely just because the builds are mapped out better than in say 30 minute 6 base games, no getting maxed, etc.)
Also, theoretically it would be possible to account for the differences in macro mechanics between races but the analysis would be quite computationally expensive and a real pain in the ass. Not sure if it'd be possible.
|
This was a great write up. Thanks so much for the read.
|
TL, Y U no make one for the Latin America ladder?
|
On October 31 2011 01:34 MotorDouglas wrote: TL, Y U no make one for the Latin America ladder?
It's combined into one "Americas" server with North and South America.
|
Canada1169 Posts
DeMu out macro-ing Bomber :o
|
|
|
Great article, I learned a lot. It's fun to see the pros compared in different ways.
I agree with what some have said, I think the warp-in mechanic will have inherently lower SQ for Protoss because the unspent resources will be higher in-between warp-ins. I'm sure that's something you'll look into.
|
Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all.
|
this is amazing yet again
i imagine SQ numbers can really depend the player's style however... the builds, and mistakes in small lapses where someone macros poorly for a couple minutes. i woould personally like to see average game length also! and maybe the matchups at some point in time
|
One of the most interesting TL posts i've read in sooooo long, especially since it's incorporates what i've learned in my stats class :D:D
good job! would really like to see data for other servers particularly the korean server
|
excellent. thank you for all the work in pulling together this excellent and intriguing analysis.
|
Awesome write-up, very impressive.
If I may, you shouldn't compare players of different races to each other because because each race's economies function too differently. For example, a Protoss with 10 gateways late game is always going to float a lot of money between warp ins whereas Terran's barracks should be constantly queued, thus keeping their money consitently low. You'll notice this reason might explain why the Protoss in your study are scoring lower than the others.
|
very nice, and also interested in more comparisons, for example korea yes.
Btw, while we all know some just abuse apm, even on top level there are the so called useless clicks, do you think there is some way to link apm in 'real' battles to some sort of overall performance, and perhaps as some kind of minor indicator?
and i agree with the post just before me, i think you cant just compare races crosswise, since different ways of spending apply.
|
This implies that all builds are fine-tuned to never have money floating. Although this is more or less the case for Terran and Protoss, Zerg doesn't follow this MO. The most common ZvT build, 2-3 hatch Muta, usually involves periods of time when gas and minerals are floating anywhere from 600 to 1000 of both minerals and gas. This isn't bad play, but the optimal method of having a relevant amount of Mutalisks.
The same can be said for making Broodlords, corrupters in ZvP, and Infestor builds.
|
On October 30 2011 20:21 ikona wrote: Doesnt such small sample size defeat the purplse of this analisys though?
I mean you can get different sets of last 30 games from the same player that provide wildly different results all day.
So what is this data for exactly?
So you're talking about sampling error, which is a fair criticism. We have some evidence, though, to suggest that sampling error is not a major factor in these results.
Take the Euro/US players. That both of those players' SQ values were not significantly changed between their US 30 and their Euro 30, suggests that the SQ values don't vary greatly from 30-game set to 30-game set. Now that's still only 60 games, but the lack of variation between the sets of 30 is rather suggestive.
Again, it isn't proof that the sample size is robust, but it is a good indication.
|
can somebody clearify?
SQ = average income/average unspent minerals
am I correct?
|
On October 31 2011 10:37 aztrorisk wrote: can somebody clearify?
SQ = average income/average unspent minerals
am I correct?
No, read the original thread and there is an exact equation for it.
|
This is truly an impressive post. Interesting to see.
|
On October 30 2011 15:59 tgun wrote:Show nested quote +On October 30 2011 12:32 JeffVader wrote:Wouldn't Terran's SQ naturally be higher simply because they can queue units up in their production facilities? I'm not trying to take away from this excellent analysis by the way . If you're aiming to purely increase your SQ then yes; however, since everyone in Ladder games (at least at the top level) isn't queing things up or spending just for the sake of spending, it shouldn't be a factor.
Thanks. Very good point.
|
Such a good article, more statistics but this one is specially interesting, gives you a lot to think of.
|
very interesting article and findings. I would love to see more from you man!
|
DeMuslim :O very impressive
|
On October 31 2011 07:27 KurtSib wrote: Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all.
Because when one doesn't have a clear assumption that it is okay to use a 1-tailed t-test one should always use the more "foolproof/safe" 2-tailed test. Also he provides p-values, which is all that is needed to asses whether or not H0 can be supported.
|
cant wait for these results!
|
It's nice to see my average SQ close to Sheth's, my min SQ in low Masters and my high SQ about Ret's average. Now I know I don't need to macro better and instead I should actually start learning some timings and maybe to control my units...
|
Amazing write-up! Hope to see more soon!
|
I just made an account just to say this is an awesome report!
|
On October 31 2011 18:07 Grovbolle wrote:Show nested quote +On October 31 2011 07:27 KurtSib wrote: Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all. Because when one doesn't have a clear assumption that it is okay to use a 1-tailed t-test one should always use the more "foolproof/safe" 2-tailed test. Also he provides p-values, which is all that is needed to asses whether or not H0 can be supported.
Why doesn't that assumption exist for the Idra ladder vs. Idra tournament comparison?
|
I may have missed it earlier in the comments but has anyone explored whether SQ is a good predictor for the outcome of a game? There are clearly limitations to using it as a prescriptive indicator given that it doesn't take into account micro or tactical/strategic decision making but I'd be interested in seeing how frequently the person with the higher SQ wins the match. Doing this and observing differences in results in different match ups could also give evidence as to whether SQ is an equal indicator of skill across all races (If TvT the higher SQ player wins 67% of the time, PvP they win 53% of the time it's probably a more useful metric for judging Terrans etc). More data is probably needed to do this but I think looking at SQ in head-to-head match-ups could help people determine how important SQ is.
|
Can't wait to see the Korean stats.
|
Someone should figure out how to pull this data from a replay. I haven't fiddled around with reps myslef, but I'm sure it's possible with enought time.
|
This math gave me a headache.
|
On October 29 2011 18:58 zul wrote:GoODy in front of IdrA in a Statistic about macro abilitys. Who would have thought that
Just shows that queuing is an incredible tool to keep up your macro while yo u need to attend to other things.
Also a good explanation why terrans tend to gave the best SQs, and protoss the worst:
All terrran production is queued, so it is easiest for them to have constant production and cushion slip-ups.
Most protoss production is from Warp-ins. Not only can't it be queued, it also requires direct vision and attention on the battlefield and sometimes rewards not spending and build up warpgate count.
The zerg larva production is in between in how much player attention and involvement it requires. While there aren't queues to keep up the production times for you, stockpiling larva can make up for previous slip-ups to a degree.
In short, it is easiest to be efficient with terran production, as long as you know when, what and how many structures to build. The efficient automatic use and allocation of production slots more than makes up for temporarily tying resources into the queue. Terran production is most forgiving with the timing of the production button presses.
|
Last 20 laddergames I had 88. I guess shorter games makes it easier to get high SQ?
|
Interesting that the max seems to be 116 so far. Is there anything special in regards to that result?
|
A very nice write-up. Just goes to show how smart people in our community are. As starcraft 2 develops more and we are able to come up with statistics which have very close correlation to the performances of the players, it will greatly help the development of E-Sports in the same sense that statistics helped sports such as basketball (NBA), baseball (MLB), cricket, swimming, etc. It would be great if possible that somehow Blizzard is able to incorporate further statistics into the game to help players gauge their progress and people who watch and analyse starcraft a quantitative measure of differences between the skill levels of different players.
There has also been lots of post commenting on how different races have different macro mechanics and that different play styles that is unique to their own race making the SQ inaccurate. My view is that as of now the idea of SQ is a great one and in its early stage my take some tweaking to the way its calculated to provide an absolute quotient for comparison across all races. The way the SQ is right now still works nicely, in that it should be used as a comparison between players of the same race with some sort of benchmark. For example, in baseball you would have your pitcher's whose skill level is rated by stats such as 'ERA' or '# of strikeouts', whereas a batter's value to its team would be measured by stats such as 'HR', 'RBI', etc. Perhaps a more relevant analogy would be in basketball, where all the players are measured with the same stats such as points, assists, rebounds, steals, blocks, turnovers, etc. but certain statistics are valued higher for certain positions. For example, assists, steals and points are more relevant to a point guards but for the center blocks and rebounds are more relevant. It is not useful to compare a point guard's stats with a center's stats, but instead it is more useful to compare point guard's stats are against each other. In the same sense, it would not be useful to compare a terran player's stats with a zerg player's stats.
Sorry for the long winded explanation, but that's my 2 cents on the discussion.
|
On November 01 2011 10:00 SnowK wrote: Interesting that the max seems to be 116 so far. Is there anything special in regards to that result?
"...and an astounding 121(!) by DroneKing (LiquidRet)."
|
On November 01 2011 12:10 OminouS wrote:Show nested quote +On November 01 2011 10:00 SnowK wrote: Interesting that the max seems to be 116 so far. Is there anything special in regards to that result? "...and an astounding 121(!) by DroneKing (LiquidRet)."
Getting anywhere above 100 is taxing as hell, and usually detrimental to overall gameplay. I played a game where I, essentially, only worried about my SQ. High drone count, low minerals at all times. You then realise how many times you will stockpile minerals to pop out a bunch of a new unit: infestors, as your research ticks over. You stockpile money to pop a large amount of banes in case of an attack, etc. I ended the game with a 118, but it was one of the harder victories I've had in a long time, purely due to the fact that I never floated money to pop a large amount of tech, and had more drones than I'm usually comfortable with. Your SQ will not be too high, unless your drone count is high, due to the way it's calculated.
|
Great article it seems you really put a lot of work into it. It makes sense that IdrA had an SQ of 96 at MLG but i still don't understand why Huk scored comparatively low at 79
|
On November 01 2011 13:00 Mutality wrote: Great article it seems you really put a lot of work into it. It makes sense that IdrA had an SQ of 96 at MLG but i still don't understand why Huk scored comparatively low at 79 Well he mentioned the protoss in MLG in general scored lower SQ.
|
On November 01 2011 06:49 imbecile wrote:Show nested quote +On October 29 2011 18:58 zul wrote:GoODy in front of IdrA in a Statistic about macro abilitys. Who would have thought that Just shows that queuing is an incredible tool to keep up your macro while yo u need to attend to other things. Also a good explanation why terrans tend to gave the best SQs, and protoss the worst: All terrran production is queued, so it is easiest for them to have constant production and cushion slip-ups. Most protoss production is from Warp-ins. Not only can't it be queued, it also requires direct vision and attention on the battlefield and sometimes rewards not spending and build up warpgate count. The zerg larva production is in between in how much player attention and involvement it requires. While there aren't queues to keep up the production times for you, stockpiling larva can make up for previous slip-ups to a degree. In short, it is easiest to be efficient with terran production, as long as you know when, what and how many structures to build. The efficient automatic use and allocation of production slots more than makes up for temporarily tying resources into the queue. Terran production is most forgiving with the timing of the production button presses.
That and idra doesn't give a damn about ladder. His avg of 96 SQ in an MLG shows that clearly.
|
On November 01 2011 03:16 KurtSib wrote:Show nested quote +On October 31 2011 18:07 Grovbolle wrote:On October 31 2011 07:27 KurtSib wrote: Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all. Because when one doesn't have a clear assumption that it is okay to use a 1-tailed t-test one should always use the more "foolproof/safe" 2-tailed test. Also he provides p-values, which is all that is needed to asses whether or not H0 can be supported. Why doesn't that assumption exist for the Idra ladder vs. Idra tournament comparison?
One could argue that it does, one could also argue that IdrA is better when laddering because there is no pressure. It is based on the assumptions whether or not to use a 1.tailed t-test. But it is ALWAYS (99%) okay to just use a 2.tailed t-test
The conclusion that the difference in the means are statistically significant is correct but not really really strong (I assume he chose a 5% alpha level) because p-value<alpha but not by more than 0.03. So yes, a t-value would make it so you could test whether or not this also was the case with 1.tailed t-test.
It is however ironic that if you assume IdrA is playing equally in ladder and tournament, you can conclude he doesn't. However if you assume he is better in tournaments, you are probably not gonna come to the same conclusion
|
On October 31 2011 22:14 ForgottenOne wrote: It's nice to see my average SQ close to Sheth's, my min SQ in low Masters and my high SQ about Ret's average. Now I know I don't need to macro better and instead I should actually start learning some timings and maybe to control my units... Haha, same. I'm high diamond zerg now and my SQ avg for last 30 games is 83, a high of 97 and low of 67, with avg income of 1593 over these 30 games.
|
On November 01 2011 20:39 Grovbolle wrote:Show nested quote +On November 01 2011 03:16 KurtSib wrote:On October 31 2011 18:07 Grovbolle wrote:On October 31 2011 07:27 KurtSib wrote: Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all. Because when one doesn't have a clear assumption that it is okay to use a 1-tailed t-test one should always use the more "foolproof/safe" 2-tailed test. Also he provides p-values, which is all that is needed to asses whether or not H0 can be supported. Why doesn't that assumption exist for the Idra ladder vs. Idra tournament comparison? One could argue that it does, one could also argue that IdrA is better when laddering because there is no pressure. It is based on the assumptions whether or not to use a 1.tailed t-test. But it is ALWAYS (99%) okay to just use a 2.tailed t-test The conclusion that the difference in the means are statistically significant is correct but not really really strong (I assume he chose a 5% alpha level) because p-value<alpha but not by more than 0.03. So yes, a t-value would make it so you could test whether or not this also was the case with 1.tailed t-test. It is however ironic that if you assume IdrA is playing equally in ladder and tournament, you can conclude he doesn't. However if you assume he is better in tournaments, you are probably not gonna come to the same conclusion
Very cool. Thanks for the help.
|
Awesome! looking forward to see asian data
|
This is phenomenal, my friend and I have been using your SQ system for almost a week now. Its a really great motivational tool because you can lose a game and still be able to say, "I had pretty good macro that game, better take a look at my micro and decision making." Thanks WhatTheFat, Can't wait to see what it looks like with data from more tournaments and regions!!!
|
On November 01 2011 14:24 89andy wrote:Show nested quote +On November 01 2011 06:49 imbecile wrote:On October 29 2011 18:58 zul wrote:GoODy in front of IdrA in a Statistic about macro abilitys. Who would have thought that Just shows that queuing is an incredible tool to keep up your macro while yo u need to attend to other things. Also a good explanation why terrans tend to gave the best SQs, and protoss the worst: All terrran production is queued, so it is easiest for them to have constant production and cushion slip-ups. Most protoss production is from Warp-ins. Not only can't it be queued, it also requires direct vision and attention on the battlefield and sometimes rewards not spending and build up warpgate count. The zerg larva production is in between in how much player attention and involvement it requires. While there aren't queues to keep up the production times for you, stockpiling larva can make up for previous slip-ups to a degree. In short, it is easiest to be efficient with terran production, as long as you know when, what and how many structures to build. The efficient automatic use and allocation of production slots more than makes up for temporarily tying resources into the queue. Terran production is most forgiving with the timing of the production button presses. That and idra doesn't give a damn about ladder. His avg of 96 SQ in an MLG shows that clearly.
Not even mentioning that money in queues actually hides inefficient spending from the SQ.
|
On November 02 2011 01:31 KurtSib wrote:Show nested quote +On November 01 2011 20:39 Grovbolle wrote:On November 01 2011 03:16 KurtSib wrote:On October 31 2011 18:07 Grovbolle wrote:On October 31 2011 07:27 KurtSib wrote: Why no t-values?
I'm not totally versed in stats. My major, bio, mostly flirts with them. Still, I wonder why did you choose exclusively two-tailed t-tests?
As I say, I'm far from expert, so take my questions as questions if you take them at all. Because when one doesn't have a clear assumption that it is okay to use a 1-tailed t-test one should always use the more "foolproof/safe" 2-tailed test. Also he provides p-values, which is all that is needed to asses whether or not H0 can be supported. Why doesn't that assumption exist for the Idra ladder vs. Idra tournament comparison? One could argue that it does, one could also argue that IdrA is better when laddering because there is no pressure. It is based on the assumptions whether or not to use a 1.tailed t-test. But it is ALWAYS (99%) okay to just use a 2.tailed t-test The conclusion that the difference in the means are statistically significant is correct but not really really strong (I assume he chose a 5% alpha level) because p-value<alpha but not by more than 0.03. So yes, a t-value would make it so you could test whether or not this also was the case with 1.tailed t-test. It is however ironic that if you assume IdrA is playing equally in ladder and tournament, you can conclude he doesn't. However if you assume he is better in tournaments, you are probably not gonna come to the same conclusion Very cool. Thanks for the help.
After revising my answer I see that the bolded paragraph is total bullshit, IF you assume that IdrA is better at tournaments you need LESS signifcant results to conclude that he is better at tournaments. So using a 1.tailed t-test would give the same conclusion as a 2.tailed t-test in this example. sorry for the confusion
|
I don't quite understand how SQ is calculated using the average unspent resources and the average income. It isnt a direct proportion. Maybe this was in the previous article. Anyway it would be nice to be able to calculate this for my own games.
|
On November 02 2011 23:26 Insomni7 wrote: I don't quite understand how SQ is calculated using the average unspent resources and the average income. It isnt a direct proportion. Maybe this was in the previous article. Anyway it would be nice to be able to calculate this for my own games. Don't be an ass, it is not everyone else's job to read for you and then spoonfeed the information to you.
His first article talks in detail about how the calculation is done. There are several links to it in the original post of this thread, yet there are people like you asking for it on every page.
IMO TL should ban people for making posts that clearly show they did not read the thread before posting. Some forums have this policy and it makes it a lot more worthwhile to read beyond the original post. The original post of this thread is so well thought out and written, and the discussion goes to shit before the end of the first page because moderator expectations are so low here. Sorry to go on a rant but I just think it's really too bad how real discussion on TL is so buried under pages of posts like the one I quoted.
|
Demuslim is also owning it up in the SQ leaderboards! Haha! Or he's a queue-ing whore! Unlikely, but that's be funny!
|
On November 03 2011 08:22 Vladimir wrote:Show nested quote +On November 02 2011 23:26 Insomni7 wrote: I don't quite understand how SQ is calculated using the average unspent resources and the average income. It isnt a direct proportion. Maybe this was in the previous article. Anyway it would be nice to be able to calculate this for my own games. Don't be an ass, it is not everyone else's job to read for you and then spoonfeed the information to you. His first article talks in detail about how the calculation is done. There are several links to it in the original post of this thread, yet there are people like you asking for it on every page. IMO TL should ban people for making posts that clearly show they did not read the thread before posting. Some forums have this policy and it makes it a lot more worthwhile to read beyond the original post. The original post of this thread is so well thought out and written, and the discussion goes to shit before the end of the first page because moderator expectations are so low here. Sorry to go on a rant but I just think it's really too bad how real discussion on TL is so buried under pages of posts like the one I quoted.
If you think that this post does anymore for the forum than mine, than you would be mistaken. Frankly I wouldn't mind seeing people banned for making offtopic rants in the middle of discussions about the original post. If you look there is actually nothing in the original post about how this value was calculated. Ill take a look at the first post, but calm down buddy, you sound mad.
|
What about SEA, Korean ladder and the others?
|
On November 04 2011 04:11 nucLeaRTV wrote: What about SEA, Korean ladder and the others? Coming soon!
|
huk troll face at the end made this article go from great to fantastic
|
|
i KNEW faith was the best macro terran in north america! =D amazing player...if he ever starts playing a lot again, expect some tournament results.
|
Thank you for this analysis. Makes me proud to have people like this working in the community.
|
On November 04 2011 06:41 whatthefat wrote:Show nested quote +On November 04 2011 04:11 nucLeaRTV wrote: What about SEA, Korean ladder and the others? Coming soon! I for one am waiting with great interest!
|
On November 06 2011 05:54 Vladimir wrote:Show nested quote +On November 04 2011 06:41 whatthefat wrote:On November 04 2011 04:11 nucLeaRTV wrote: What about SEA, Korean ladder and the others? Coming soon! I for one am waiting with great interest!
Me too. btw the sq calculator site (http://beijers.eu/share/sq.html) seems to be down for me atm.
UPDATE: sAsThark made a downloadable sq calculator! here it is: http://www.sas-team.net/download/file.php?id=1281
|
This is really interesting, can't wait for the other regions!
|
On November 05 2011 22:40 LtCalley wrote: i KNEW faith was the best macro terran in north america! =D amazing player...if he ever starts playing a lot again, expect some tournament results.
Thanks =)
|
Holy crap that is a lot of hard work computing all this. I have half a bachelors in actuarial math (i.e. a shit-ton of stats to study) and I love math, but I don't think I love it enough to be willing to compute all that lol. Great job man!
|
Would implementing z-scored and t-scored help lower the discrepancy between the races? Especially because of the spending styles?
|
Hey whatthefat, is it possible that you calculate the SQ of MLG Providence?
|
Thank you very much for you research and information. Understanding and putting into practice this data has provided me with a quantifiable method to track my training, which has been tremendously helpful. Again, thank you for your work.
|
I hope it's okay to bump this but I wanted to know if any more work had been done on this. I'm rather interested in evaluating players with standards and statistics outside of wins and losses and this seems like a very promising way to do so. One of the interesting things about other sports is being able to compare statistics between players but in Starcraft 2 there aren't that many ways to do that. For example, in basketball you can compare who gets more assists per game or what their player efficiency rating is. Starcraft 2 woud benefit from statistics such as these.
|
|
|
|