|
wee terrible statistics at work...
This is just so bad it's laughable and the 'conclusions' of it are terrible. For a huge replay pack this could be somewhat useful but for anything less then a couple thousand replays this is just bogus, especially when using this method.
Especially the OP's comment to discuss 6-pool every time or 2 rax every time is stupid. I know it's most likely meant as a joke, at least I hope it is, but in the light of the terrible graphs I fear he actually thinks these graphs mean anything.
User was temp banned for this post.
|
On February 06 2012 06:04 Markwerf wrote:
wee terrible statistics at work...
This is just so bad it's laughable and the 'conclusions' of it are terrible. For a huge replay pack this could be somewhat useful but for anything less then a couple thousand replays this is just bogus, especially when using this method.
Especially the OP's comment to discuss 6-pool every time or 2 rax every time is stupid. I know it's most likely meant as a joke, at least I hope it is, but in the light of the terrible graphs I fear he actually thinks these graphs mean anything.
On February 05 2012 06:22 ZeroTalent wrote:
The data does show some interesting results. But I'm not satisfied that the sample size is large enough. I'd love to repeat this analysis on a bigger dataset if there is fresher and larger replay pack lying around (coughplayhemcough).
To the Pros, Masters players, and everyone: do the numbers here line up with how you think about the game? Is Terran overmatched in the late game? Does Zerg grow stronger as the game goes on? Should you 6-pool (100% Z win rate in PvZ <5 minute games) or 2-rax every (80% T win rate in ZvT 5-10 minutes) every game? Discuss.
OP discusses the data, says he's not satisfied with the sample size and points out the ridiculousness of the conclusions which can be drawn from his data, and still gets flamed for "think[ing] these graphs mean anything".
I think the OP brings up a good point, that analyzing win rate match up data and game length may yield novel conclusions about the nature of each match up. When you design an experimental protocol, you don't just go and run your experiment on the whole population you are testing, because your procedure may be total shit. Usually its a good idea to test if the protocol yields reasonable results on just a small sample size. This is basically what the OP is doing, and I would say some of the things shown in the data are indicative (such as the early strength of terran against zerg) that this method of analysis could yield something which reflects the nature of each match up.
So yes, these graphs do mean something, just not what you think.
|
yep late game terran doesn't look so strong, while early game terran is beast. Maybe this is why every race's best strategy against terran is defend and macro until late game
|
Huge props to the author for sitting down with the dataset and doing the number crunching! This is really interesting, although the conclusions that can be drawn are limited by the particulars of the data.
I'd love to see this done with either a bigger or more specialized dataset. As others have noted, many matches (if the dataset is indeed all of Providence including open bracket) will have great disparities in skill, but I still think this gives some (very uncertain) sort of indication of how the various races fare in different situations. The same analysis performed with (say) all top 32 MLG games, or all => ro32 Code S games could yield in some very interesting statistics, especially if graphs are provided for "all of history" as well as grouped by patch/metagame shifts. Gathering the data (the most difficult/dreary part, I reckon) and performing the analysis would - of course - take a lot of time and perhaps not be a lot of fun, but I think the results could be illuminating.
Again, nice work!
|
I am pretty sure the average game length of pvz is only at 10 mins.
that's about when the 2 base all-in starts to hit
cam't remember seeing a macro toss at all
|
Interesting, I wonder if it would be even more interesting if we could look at the specific matchups :D
|
On February 06 2012 12:57 ETisME wrote:
I am pretty sure the average game length of pvz is only at 10 mins.
that's about when the 2 base all-in starts to hit
cam't remember seeing a macro toss at all
Between the instant tech switching, ling runbys and out of control macro, why would any Protoss in a PvZ choose for a long macro lategame?
I didn't expect 15-20min TvP to be so wildly in T's favour, I had thought that 1 base allin timings to be generally over and that the double chronoed forges and T3 to kick in right about that point.
|
On February 05 2012 08:37 jupiter6 wrote:Show nested quote +On February 05 2012 08:29 Aquila- wrote:
I knew it, Terran gets ridiculously hard when the game goes on longer. It is almost impossible to win a long macro game on maps like Taldarim, but Terrans still keep their winrate at 50% with allins or 1 base timings. I wish Terran would get a stronger lategame, then we would see less allins and more macro games. I mean why would you play a long game if you know your chances of winning get smaller with the time?... this imo is design flaw, i love terran but its so annoying playing this race is like playing against clock
This This This. Terrans late game is weak compared to Toss and Zergs. What viable tech option can terrans look to tech to thats actually viable? A barracks unit called a ghost lol???
If ravens were less expensive and had a longer range so that you didnt have to suicide them when using their abilitys they'd actually be viable. If BC's had an ability simliar to the carriers and BL's that shoots out units at a long range they'd actually be viable. Blizzard should make BC's shoot out parachuted marines that have combat shields from a range of 9.
To the people that talk about terrans winning in the late game, the only time that a terran will win in the late game is if their micro is far superior to their opponents. Whens the last time that you saw someome baneling/ling splitting just as well as MVP's marine splits and stutter steps?
|
On February 05 2012 10:00 Greenei wrote:
this is pretty much excatly what i as a terran master would expect.
i checked 800 games with sc2gears:
5-10 min 66%
10-15 min 61%
15-20 min 64%
20+ min 42%
The problem is this could be just as indicative of your playstyle or relative strengths as it might say anything about the race in general.
|
On February 06 2012 12:57 ETisME wrote:
I am pretty sure the average game length of pvz is only at 10 mins.
that's about when the 2 base all-in starts to hit
cam't remember seeing a macro toss at all
Wasn't Hero vs Ret a long game that ended with a vortex or two? I'm pretty sure that was Providence.
I think you'll find more than a few long PvZ games, just protoss loses most of them. 
On February 06 2012 13:34 Sovern wrote:Show nested quote +On February 05 2012 08:37 jupiter6 wrote:On February 05 2012 08:29 Aquila- wrote:
I knew it, Terran gets ridiculously hard when the game goes on longer. It is almost impossible to win a long macro game on maps like Taldarim, but Terrans still keep their winrate at 50% with allins or 1 base timings. I wish Terran would get a stronger lategame, then we would see less allins and more macro games. I mean why would you play a long game if you know your chances of winning get smaller with the time?... this imo is design flaw, i love terran but its so annoying playing this race is like playing against clock This This This. Terrans late game is weak compared to Toss and Zergs. What viable tech option can terrans look to tech to thats actually viable? A barracks unit called a ghost lol??? If ravens were less expensive and had a longer range so that you didnt have to suicide them when using their abilitys they'd actually be viable. If BC's had an ability simliar to the carriers and BL's that shoots out units at a long range they'd actually be viable. Blizzard should make BC's shoot out parachuted marines that have combat shields from a range of 9. To the people that talk about terrans winning in the late game, the only time that a terran will win in the late game is if their micro is far superior to their opponents. Whens the last time that you saw someome baneling/ling splitting just as well as MVP's marine splits and stutter steps?
This is pretty funny, since most people will declare a TvZ pretty much over if the Terran gets decent positioning with 16+ ghosts. Unless it's fin vs leenock
I do agree that thor / BC are not great 'late game' tech, but for almost all races tier 3 sucks. Colossus are good, more like essential, with Broods and Ultras coming in after that. Then Thors as long as something is flying somewhere in the game. BCs bringing up the rear with the 'mighty' Carrier right at the bottom.
To be fair to the terran race, their mid game units are the strongest in the game. MMMGV is a good counter to the Protoss death ball, and marine, tank, thor(if muta), viking(if BL), ghost(if anything?) does well vs zerg. Most of the time it looks to me like the player who is playing better on any given day wins more than racial imbalance. I would argue that Terran have a kill window where the other two races are trying to kick their tier 3 into gear while terran already have their 'strongest' composition and this is where Terran aim to make the game end.
|
I think in a very very very general sense terran is the most lethal race earliest on in the hands of skilled players
|
Always interesting to be able to take in such information in this fashion.. Don't know that too many conclusions can be drawn from it though.
Thanks for putting it together!
|
MLG providence, too few data points.
Also, I'd really like a breakdown of when they won in the game. This is sort of that, but not quite what I want.
TvP AND PvT, who's winning at which points of the game
ZVT vs TVZ, etc...
|
On February 06 2012 17:34 Divinek wrote:
I think in a very very very general sense terran is the most lethal race earliest on in the hands of skilled players While that might be true, I think another big factor is that terran is by far the best defensive race (even though they're good offensively as well), meaning if they play smart will win a majority of their games where the opponent does any sort of commitment to an early-game attack.
I think it's really a combination of them both. Due to the mechanics of zerg, they have probably the strongest early on potential, I'd say (most efficient t1 unit for the early game, and banelings can destroy anything that can counter the zerglings, such as walls). Thing is, that doesn't give zergs necessarily the highest win-rate for short games, since they might not all play like that, and/or their tactics can easily be dealt with when scouted. For instance, a player who doesn't run aggressive all-in short-term builds will oftentimes succumb to another build attacking offensively early on.
Anyway, whatever; I brought up too much theory craft than I care to.
|
This is exactly why I mainly all-in mid- or earlygame in TvZ and TvP and only play Macro games in TvT.
I'd say Terran is overpowered in the early and midgame but extremely underpowered lategame and about ten times harder to play in a macro game than either Zerg or Protoss. In TvT I know that I won't lose to inferior players in a macro game just because their race is easier to play late game and has a generally stronger late-game aswell.
And if people want to flame me for that, read what cloud said on page 2 (?), he said essentially the same thing, that terran is easily the hardest race to play in a macro game, hence why they have the lowest win rates in 20+ mins game (and because imo terran lategame is underpowered, but like I said, it's balanced out because early and midgame terran is OP, but I feel like if the toss or zerg just defend and drag the game into the lategame I can't win with terran)
|
On February 06 2012 12:27 RedMosquito wrote:
yep late game terran doesn't look so strong, while early game terran is beast. Maybe this is why every race's best strategy against terran is defend and macro until late game
Or Terran is more dominant in the early stages than in the later stages. Thus a greater share of its victories are in the early stages. Toss on the other hand may be less effective in the early stages and enjoy more of its total share of victories in the late stage.
The graphs point to T strength in the early stage but do not necessarily say anything about T in the late game. We're talking about each individual race's share of victories.
|
On February 08 2012 11:47 Sabu113 wrote:Show nested quote +On February 06 2012 12:27 RedMosquito wrote:
yep late game terran doesn't look so strong, while early game terran is beast. Maybe this is why every race's best strategy against terran is defend and macro until late game Or Terran is more dominant in the early stages than in the later stages. Thus a greater share of its victories are in the early stages. Toss on the other hand may be less effective in the early stages and enjoy more of its total share of victories in the late stage. The graphs point to T strength in the early stage but do not necessarily say anything about T in the late game. We're talking about each individual race's share of victories.
??????????????????
These are percentage graphs.
|
Pretty cool graphs, I always find stuff like this interesting. One big thing I think is you shouldn't have the overall win rate on the end of the graph. It doesn't make sense there, and ruins the flow of the line. Other than that though it's well done. Obviously as some people pointed out a bit longer time frame would be a good thing to add if you managed to get more replays so you had a large enough sample for extended games. A big replay pack release would be really nice for that.
|
I don't think it has anything to do with the race, but it has to do with the player
The probability of the better player winning approach 1 the longer the game goes on. Just look at how cheesy players get wins (versus time) and how players like Flash win
|
On February 08 2012 12:07 SkimGuy wrote:
I don't think it has anything to do with the race, but it has to do with the player
The probability of the better player winning approach 1 the longer the game goes on. Just look at how cheesy players get wins (versus time) and how players like Flash win
You are right, it's not the race, all terrans are just retarded and can't macro.
|
|
|
|
|
|
EPT
