EPTYet the exponential relationship between income and unspent resources holds up, even at very high income levels. It looks like it may begin to deviate slightly for income > 2000. In other words: sure, Grandmaster players start to accumulate lots of money in these situations, but they are still spending it much better than lower league players. This is probably largely due to stronger players thinking of clever ways to keep spending to try to improve their position even when maxed, e.g., making efficient army trades, sacrificing units to make tech switches, etc.
Do you macro like a pro? - Page 20
| Forum Index > SC2 General |
|
whatthefat
United States918 Posts
Yet the exponential relationship between income and unspent resources holds up, even at very high income levels. It looks like it may begin to deviate slightly for income > 2000. In other words: sure, Grandmaster players start to accumulate lots of money in these situations, but they are still spending it much better than lower league players. This is probably largely due to stronger players thinking of clever ways to keep spending to try to improve their position even when maxed, e.g., making efficient army trades, sacrificing units to make tech switches, etc. | ||
|
Daozzt
United States1263 Posts
| ||
|
Surrealz
United States449 Posts
It would be interesting to try and integrate SQ into other parameter tests and see if there are any relationships there as well. My only complaint is the small sample size of 100 games per, which is really really low for the kind of claims you are making. If it was 1000 games per I would be much more confident in these results. | ||
|
sashamunguia
Mexico423 Posts
:D yeah!!!!, finally some ppl bringing math into SC2 you did a pretty nice representation of data. We all know it can be improved a lot and start considering many other tiny factors perhaps, but this is a huge step forwards really. I cant wait to see teams having their own charts measuring their progress, or some other nerds polishing your equations even further. Suggestions: --- What about taking in account differences not only in races, but for example, in matches, i.e. would a terran macro as hard playing against a zerg as he does vs another terran? --- When taken in account the games to make the averages and shit, it may be nice (If I do my games) to exclude "awkward" games, like the ones in which i lost 40 SCVs to BFH. Or taking a loot at the other side of the coin, what about (again, me myself of my games) making a separate chart for this kind of "awkward" games, to measure how good was my macro-reactin after losing a buckload of SCVs. --- Talking to MLG, SC2CHARTS , or TeamLiquid, whoever in order to establish a database, and compare the macroing abilities of different players. SC2CHARTS Is a very good place to incorporate this IMO :D, and well, MLG releases his own numbers after each event (number of motherships used and shit hahah but they might as well use their numbers for something better  ) | ||
|
theBOOCH
United States832 Posts
| ||
|
Reclus
Norway1 Post
Could you please post the coefficients and p-values, as I would like to be able to compare my worker production to the different leagues. It would also be awesome if you could test if the difference between the leagues are significant, and on what levels. Thanks for the entertaining read! | ||
|
betaflame
175 Posts
| ||
|
TheAmazombie
United States3714 Posts
| ||
|
stanik
Canada213 Posts
Games ending 11-20 min (4 games) - SQ = 97.4 Games ending 20-30 min (13 games) - SQ = 88.8 Average SQ of 92.3 Avg SQ in Wins 93.3 Avg SQ in Loss 91.2 Average Unspent Resources - 609 Average Resource Collection - 1444 I am a low Masters Terran | ||
|
Clbull
United Kingdom1439 Posts
On September 17 2011 05:00 kmh wrote: Guess what? Your results fall into the expected range for a gold league player regardless of that inaccuracy. Does that matter? It was still significantly lower than it should have been regardless. 8 points less is a big fucking deal. | ||
|
Horse...falcon
United States1851 Posts
| ||
|
KimJongChill
United States6429 Posts
| ||
|
stanik
Canada213 Posts
On September 17 2011 03:28 Grobyc wrote: Woah, the difference in workers produced between Grandmasters and Masters is actually a lot more than I expected. Nice work and easy to understand, thanks whatthefat I think much of that difference comes from Grandmasters having to replace workers that are lost from superior harassment from opponents. or maybe masters players forgetting to replace lost workers. | ||
|
Ownos
United States2147 Posts
On September 17 2011 05:00 kmh wrote: Guess what? Your results fall into the expected range for a gold league player regardless of that inaccuracy. Haha you're right. I think people need to sample more of their own games instead of taking a small range with several odd ball games that could throw it off before they can correlate SQ with skill. And that's also probably assuming you're playing to win. What would happen if someone just cheesed 100% of his games and had high SQ. That's why I say, people should categorize their data into time frames. Anyone who follows a build order closely will have high SQ. It doesn't mean much if all your SQ data points are between 5-10 minutes in game length. It just means you can spend efficiently in a relatively easy time frame. We all know the longer the game, the better chance the better player will reveal itself. And I feel the mass banking / mass production games don't happen that often. Even ask yourself, is it easy for you to keep your money low in the beginning of the game? I always fall apart later in the game while pros can spend much better than me at 15 minutes. In the end, you will always have a collection of long and short games and your low SQ for long games will drag your average SQ down unless you make a strong effort to leave the game after your cheese fails and you cheese every game. That isn't much different than gold players spamming to 100+ APM. On September 17 2011 05:22 stanik wrote: Games ending 0 - 10 min (4 games) - SQ = 98.5 Games ending 11-20 min (4 games) - SQ = 97.4 Games ending 20-30 min (13 games) - SQ = 88.8 Average SQ of 92.3 Avg SQ in Wins 93.3 Avg SQ in Loss 91.2 Average Unspent Resources - 609 Average Resource Collection - 1444 I am a low Masters Terran Like what this guy did. On September 17 2011 05:24 Clbull wrote: Does that matter? It was still significantly lower than it should have been regardless. 8 points less is a big fucking deal. You can't throw out a data point just because you don't like it. It happens. Sometimes you get a data point that so far off every other data point in anything. You just need to get a larger sample. | ||
|
Amui
Canada10567 Posts
| ||
|
GhostFall
United States830 Posts
| ||
|
Ebos
United States44 Posts
Avg Unspent resource in Column A Avg Resource collection rate in Column B This formula in column C =35*(0.00137*B1 - LN(A1))+240 Then use the fill down option by selecting column c row 1 and highting multiple rows in column and clicking ctrl+d to fill down. | ||
intrigue
Washington, D.C9933 Posts
| ||
|
Ownos
United States2147 Posts
On September 17 2011 05:30 Amui wrote: How does queuing units affect average unspent resources? It seems like Terran would have a slight spending advantage in this regard compared to Protoss and zerg. I think it'd be no different than APM spamming. Most players play hoping not to queue units. Just like players hope to have a high effective APM and not make an effort to inflate it. Doing either would probably harm your game by slowing you down doing bad actions (queuing or needless spamming). I also need to note that I'm talking about spamming not tapping. | ||
|
BushidoSnipr
United States910 Posts
Just imagine y=complex equation, when y does not equal cheese | ||
| ||
