|
United States918 Posts
On September 19 2011 06:33 Excalibur_Z wrote:State brought up an interesting note about how SQ favors longer macro games if you're good at spending. Vanick constructed this graph based on SQ and it seems to be something to consider: http://www.wolframalpha.com/input/?i=plot 35(0.00137x - ln(y)) + 240, x=800..2000, y=400..1700Also I want to make sure people aren't finding their SQs and using that as some "magic number" to define their skill level like they did in that closed "calculate your MMR" fiasco of a thread. Macro is only one component of the game and there's more to it than that.
Indeed, but it becomes increasingly hard to spend well in longer macro games! The formula is designed so that a player of any particular league level is no more likely to score a high SQ for any particular income or game duration. In other words, it is normalized so that it is just as hard to score a high SQ in high income games as low income games.
Regarding what SQ represents, I completely agree that it is not some kind of magic number. I hope I conveyed that adequately in the OP. What it does represent is an objective measure of spending efficiency. Of course, there are many other components to SC2 skill, and SQ is not even a comprehensive measure of macro. Nonetheless, I think it will prove to be a very useful tool for a number of purposes, especially given the lack of such tools.
|
Amazing in-depth analysis, I love how math and SC2 comes together in an epic post like this. Thank you for your brilliant effort.!!!
|
@_@ I don't think I even understand most of what you just said. Awesome.
|
|
|
Only went as far to do my last two games, both medium length tvp's where I got 79 and 85 SQ. I'm a low-mid masters. Seems accurate, my mechanics are better than my strategy sense mostly because I can't spend the time to follow the scene.
|
Dear Whatthefat (and other TL.net readers),
Thank you for your wonderful post!
I wanted to compliment you on finding what at first appears to be a magically consistent factor among all three races in macro play.
In actuality I do not think this is random or "too good to be true."
When I've listened to Dustin Browder speak about sc2, especially in some of the earlier pre-release discussions, he always emphasized ECONOMY as the biggest thing.
He was always talking about economy!
Since he designed the game, the game reflects his biases.
Now you can say that beyond the economy of macro play is also unit composition, positioning, micro, etc, etc, which involves HOW you spend.
Your lovely consistent factor of SQ is just about THAT you spend (efficiently).
And in my research on Dustin Browder, that's what he was always talking about, at least from my experience. He was just always saying "the more resources you can acquire the better."
So Dustin is saying "the MORE you can spend (efficiently is implied) the better"
And your SQ factor says "the MORE you spend and the MORE EFFICIENTLY you spend it, the better you are.
Whatthefat:
"Of course, there are many other components to SC2 skill, and SQ is not even a comprehensive measure of macro."
The SQ is however an exact and precise measure of ECONOMY management, right? 
So truly I think your SQ factor magically balances across the 3 races because Dustin, not the only molding mind of sc2 but certainly a predominant one, DESIGNED the game to work that way.
So congratulations to you on sifting through the math and games to gather this new tool for us.
Victory to You!
-Joytoyou
PS: I posted this rather than PM-ing this because I want others to continue to join me in celebrating Whatthefat's wonderful contributions, and wanted to explain why I personally feel what he did was so wonderful. ♥
|
On September 18 2011 02:14 ScrubS wrote:Show nested quote +On September 17 2011 09:19 DiaBoLuS wrote:On September 17 2011 09:06 ScrubS wrote:
I dont get the SQ stuff, I calculated mine from 3 random games in my match history and i had an average of 93. You are suggesting that with such a high SQ number, i shouldn't focus on my macro because I am able to spent my resources? Or is it more an indicator of macro capabilities rather than your actual skill? I guess that just depends on the number of games you analyze? or your math is wrong The math wasn't wrong, besides, that isn't relevant to my question.. Show nested quote +On September 17 2011 09:29 rauk wrote:On September 17 2011 09:06 ScrubS wrote:
I dont get the SQ stuff, I calculated mine from 3 random games in my match history and i had an average of 93. You are suggesting that with such a high SQ number, i shouldn't focus on my macro because I am able to spent my resources? Or is it more an indicator of macro capabilities rather than your actual skill? I guess that just depends on the number of games you analyze? 3 is an incredibly small sample size. i can also get a 90+ SQ if i cherry pick games, that's why you need to use at least 10+ games. The sample size doesn't matter, it isn't of any importance to my question. Besides, i didnt cherry pick, when i calced it of 10 random games, my SQ was still 92... Show nested quote +On September 17 2011 09:38 tshred wrote:On September 17 2011 09:06 ScrubS wrote:
I dont get the SQ stuff, I calculated mine from 3 random games in my match history and i had an average of 93. You are suggesting that with such a high SQ number, i shouldn't focus on my macro because I am able to spent my resources? Or is it more an indicator of macro capabilities rather than your actual skill? I guess that just depends on the number of games you analyze? It's just an indicator of how well you spend your money. There are other hidden factors for macro that can't be calculated mathematically such as expansion timings, worker saturation compared to the number of bases, number of production facilities, etc. These other factors affect your macro in addition to how well you spend your money. To determine if your macro is good enough, you'll need to examine some information about your games. If you are below diamond league, then for the most part you could still work on macro. For the games that you lost, compare your income to your opponents. If you have a high SQ, but your income is lower than your opponent, then you still need to work on your macro. If your income is about the same or greater than your opponent and you still lost the game, then there was probably some decision making or micro error that might've cost you the game. Okay thank you <3 at least somebody who doesn't focus on my SQ number
omg ur sq was over 90? u must b tryin 2 brag, im goin 2 ignor ur question
On September 17 2011 21:37 Komeidon wrote:
The leagues seem to be totally off... I'm a masters terran, and it gave me 59.15 over last 27 games. I won 18 of those, so it should be a decent sample. Min: 27, max: 78.
I'm master league protoss, and my SQ is 75 over the last 10 ladder games, 82 over the last 10 custom games.
You probably just don't spend your resources very well, yet have good game sense and decision making, and the opposite is true for me.
|
On September 17 2011 20:41 Evangelist wrote:
Building in cycles does not lower the average since it is the production rate versus the amount of resources coming in that determines the average. Terran structures also tend to work in grouped cycles, so the resource distribution for them is virtually identical - it is actually more efficient to do it this way, since you keep a single internal clock and a mental count.
The macro of Protoss and Terran is actually completely identical - it just feels different. Protoss will wait for full warp ins, but not just for a sudden burst of units. It's because it is most time efficient. A terran will similarly macro in the same way, again because it is time efficient.
Peaks and troughs in a profile will be represented by the different buildings having different timings. So yes, if you've got a high unspent average, you can work on your macro. Don't excuse yourself just because you're Protoss. Your average unspent is as significant as it is for Terran, due to time averaging.
...No.
Say a protoss is waiting to warp in 4 stalkers, and his resources get up to 500 minerals (the exact cost of 4 stalkers).
Now say the protoss is producing stalkers out of an appropriate number of gateways, so he only waits till he has 125minerals before producing a stalker.
In the second case, his average unspent would be 62.5 (ignoring gas for simplicity), yet in the first case, his average unspent would be 250.
The longer you have to wait before producing units, the higher your average unspent goes.
|
Made for a fun read (i know, mathematical analyses are generally not regarded as 'fun', but I'm weird at times like that).
Appreciate the work that went into this, and the results are presented in an easy-to-absorb format.
Well done.
|
On September 19 2011 13:15 Buddhist wrote:Show nested quote +On September 17 2011 20:41 Evangelist wrote:
Building in cycles does not lower the average since it iS the production rate versus the amount of resources coming in that determines the average. Terran structures also tend to work in grouped cycles, so the resource distribution for them iS virtually identical - it iS Actually more efficient to do it this Way, since you keep a single internal clock and a mental count.
The macro of Protoss and Terran iS Actually completely identical - it just feels different. Protoss will wait for full warp ins, but not just for a sudden burst of units. It's because it iS most time efficient. A terran will similarly macro in the same Way, again because it iS time efficient.
Peaks and troughs in a profile will be represented by the different buildings having different timings. So yes, if you've got a high unspent average, you can work on your macro. Don't excuse yourself just because you're Protoss. Your average unspent iS as significant as it iS for Terran, due to time averaging. ...No. Say a protoss is waiting to warp in 4 stalkers, and his resources get up to 500 minerals (the exact cost of 4 stalkers). Now say the protoss is producing stalkers out of an appropriate number of gateways, so he only waits till he has 125minerals before producing a stalker. In the second case, his average unspent would be 62.5 (ignoring gas for simplicity), yet in the first case, his average unspent would be 250. The longer you have to wait before producing units, the higher your average unspent goes.
I agree with Buddhist, think of it like this.
There is a relationship between spending and the amount of time your units are on the map.
The longer amount of TIME your units are on the map, the more efficiently you are spending your continuous stream of income.
In cycles of production, the units have less total time on the map than if you make each unit as soon as you are able.
Just thought you might find this a useful intuitive way to understand the importance of efficiency in spending and the SQ
You want as many units around for as long as possible (whether workers or army units).
-Joytoyou
|
why is it that if your unspent is 2000 and income is 2000 you get an SQ of 69.86841391602712 which is between diamond and master. I mean if your unspent is at or above your income your SQ should be rubbish. I don't know many masters who's income is equal to unspent.
|
This thread helped me to indentify one flaw in my play: the constant worker production. I don't do it. Although as a Protoss, I sometimes don't have to ( timing attack, mirror, etc ), this seems to be where I am doing worst.
On the other hand, my average SQ is around 78, but if I only take pvt and pvz, it is more like 82. However, as a lot of ppl already said, the SQ is not really a way to find out if you macro well. It can be and indicator but if you have 80 or 85 can purely come from the choice of maps, race etc. Terrans for instance can easily queue and therefore never have high minerals, thus a very high SQ. Zerg can use their information and not produce any army until they know what they need.
Still, very useful post.
|
On September 19 2011 13:00 Joytoyou wrote:Whatthefat: "Of course, there are many other components to SC2 skill, and SQ is not even a comprehensive measure of macro." The SQ is however an exact and precise measure of ECONOMY management, right?
Depends on definition, but I think Whatthefat is correct that SQ =/= macro. It is exactly what it states to be - spending quotient; that is: normalized unspent resources with regard to income rate, so that multi-base games can be analyzed fairly on the same scale as one-base games. Note that OP goes through macro in other aspects, completely unrelated to SQ. There are even more subtle ones - for example, how well utilized all your facilities are, or is there redundancy. Simplest case: when 4-gating if you build 5 gates and that one additional gate can't be utilized that early, because you don't have the income. That's bad macro which SQ doesn't register.*
(* and no algorithmic model could register it consistently, because some builds require redundant buildings to cover a specific case which may or may not happen, so they may or may not be utilized)
|
just wow ! nice job really !
lol @ EGIdrA League btw :D
|
EPIC! This is great! Probably the only good use of having done that Statistics course lol
|
United States918 Posts
On September 19 2011 17:34 Sk3tch wrote:
why is it that if your unspent is 2000 and income is 2000 you get an SQ of 69.86841391602712 which is between diamond and master. I mean if your unspent is at or above your income your SQ should be rubbish. I don't know many masters who's income is equal to unspent.
Actually, in collecting IdrA's games, I found a few games where he had income approximately equal to unspent at around 2000. And the same went for many Grandmaster and Master level players. Once income becomes that high, it's hard for even pro players to spend it all, especially because they are likely to become maxed. Because unspent resources scale as an exponential function of income, at some point unspent must exceed income, even for the best players.
|
I dont think SQ really matters for late game zerg, as you are stockpiling minerals and gas for your instant remax....
|
I really like the idea of the SQ and I would like to see something similar added in the statistics after every game, so players can instantly see whether their macro was allright or not and whether they improved or not, but I have the feeling that firstly your SQformular is not perfect and secondly the scale you gave us is wrong too!
When reading your scale-graph it says:
masters average is about 72
GM average is like 82
and IdrA reaches 88
My problem with this is, I am a highplatin zerg in EU and my SQ in ladder is about 70-80 (which does not fit to my leagueplacement in the slightest) and when playing custom against AI I can reach 110 SQ (just tested and had a game with like 1560 income and 350 unspent).
I of course do know that custom games against AI are huge differently than laddergames, but my experiment was only to show, that the formular is either not bringing in enough variables or lacks the right way of relativizing Income and Unspent.
Other than that you did a great job with your thread and I for myself am using the SQ to check whether my macro struggled and to see improvements. So thank you for that!
|
This is by far the most detailed, user friendly, well researched post to date ive seen on SC2. Thank you for the time an dedication on this amazing post.
|
75.7566 average from Rank 1 Diamond over 20 games.
Seems quite accurate of a scale.
I've been averaging some of the top GM players SQ's and they are usually around ~89 average
|
|
|
|
|
|
EPT
