|
On September 17 2011 15:08 TheRabidDeer wrote:
Really interesting, though I still think it might be unfair towards zerg, especially zerg that favor lategame where they stockpile resources/larvae. My zerg has an SQ of 58.996 while my severely underdeveloped and weak terran has 72.440 (and one of my games actually getting a 98.093 SQ after 12 minutes :o). Or maybe I am just playing zerg wrong... I am definitely pretty passive unless its ZvT.
Do you cue up a lot of units in your production building as Terran? Afaik, this counts toward spent resources even though you are really deferring the use of the minerals and gas. Zerg can't do this.
|
On September 17 2011 17:22 kedinik wrote:Show nested quote +On September 17 2011 16:00 Ghanburighan wrote:On September 17 2011 15:49 figq wrote:On September 17 2011 14:53 Azzur wrote:
I believe there are definite gaps between leagues. The empirical data supports what many higher level players know intuitively. Bronze and GrandMaster overlap (marked area): ![[image loading]](http://i.imgur.com/QiOVf.png) Let alone the closer leagues. So we can hardly talk of gaps between leagues. There is scaling in the averages, but there is still quite a lot of variation within a league, and between different games of the same player. I don't see the problem. This is an average measure, and if you compare a 9 minute game of a bronzie where he hits a 3 rax timing attack (just over 600) to a GM micro-intensive game long game, it's much more likely for that bronzie to get a higher score on that one game. It's just a very small number of games. Also, some good players get accounts that they need to play out of bronze, so those 30ish games would contribute to the overlap too. I guarantee, 100%, that no legitimate bronze league player can out-macro a GM for even 2 solid minutes. Even during the best game of his life against a GM who is drunk off his ass, an accurate comparison of any relevant metric would show them to be night and day. The SQ just seems to loosely correlate to macro and skill without really measuring or displaying anything truly insightful, valuable or accurate.
That's not the point. A bronze-leaguer will sometimes have a fast-1-base game where his SQ is high. A GM will sometimes play a long game where their macro slips. Yes, the GM player probably macroed harder during the same time span as the bronze-leaguer, but the game did not end at that point for the GM. You are responding to a poster who carefully wrote, "if you compare a 9 minute game of a bronzie where he hits a 3 rax timing attack (just over 600) to a GM micro-intensive game long game." He did not say, "If you compare a 9 minute bronzie game to a 9 minute GM game.
|
On September 17 2011 19:57 Osmoses wrote:
How awesome would it be to have this implemented into SC2 as sort of a gauge after a game? :p
"You macroed like a Grandmaster this game!" with an arrow pointing somewhere between Idra and EmpireKas.
The Coach-AI maps (using Green Tea AI) do have this feature. I find that the graphical analysis at the end of games is hard to make sense of. But they do rank your macro by league. I'm not sure how accurate it is, but it always leaves me knowing I need to improve.
|
This was awesome.
And according to my 10 most recent ladder games that fit the criteria, I have 77 SQ. Neato.
|
|
|
Wow.
Just simply stunning work.
Great read, great work.
<3<3
|
Your teory worked perfectly analising my replays... im exactly where i suposed to be.
Nice post!!
Blizzard must integrate this to the game somehow.
|
those graphs are pretty :D
|
very nice work, very interesting read.
|
Brilliant analysis; offers some different insight into the game.
|
I loved reading the post. Great work sir.
|
The reason the overlap is so significant is because people are generally playing those with similar skill levels. Just imagine an outlier bronze with 80SQ who focuses heavily on spending money playing against an outlier GM with 80SQ who wins with strategy/multiprong micro (and hence doesnt spend as well as other GM)
Guaranteed in this matchup the GM would still sit pretty with 80SQ but the bronze macro will be taxed and he wont be spending his money nearly as well. Hence why your SQ doesn't determine your rating, its correlation, not a proof.
Edit: but still a cool correlation!
|
After having a strange case of ladder-phobia, I played 5 games and had like an 80 SQ average (88 in the last game :D), even while dropping around and moving tanks into position.
I'm in plat, I guess I just need to play more. Less lurking TL, more playing. The post above me insists that I'm still bad, though.
|
That is a goddamn amazing post.
|
On September 17 2011 01:27 DarKcS wrote:
At least this kind of proves we needed the GM league. Feel sorry for non-GM level players who had to play ladder when they were underneath them...
MMR eliminated that anyways. GM league was just made to make people feel good about themselves. As it is, GM are still playing top Masters and vice versa and the league actually doesn't make any difference whatsoever.
|
Really nice write up. Did you use excel to make the graphs or did you use a separate package? The reason why I ask is because these look a lot like the graphs you can make from ggplot.
Great read though, I love this kind of analysis.
|
On September 18 2011 13:36 evilduky666 wrote:
Really nice write up. Did you use excel to make the graphs or did you use a separate package? The reason why I ask is because these look a lot like the graphs you can make from ggplot.
Great read though, I love this kind of analysis.
He said he used Mathlab
Good (and lots of) work here, gratz !
(we might need to add alcohol as a variable somewhere, just in case a little bit of it makes you more aware.)
|
On September 18 2011 11:31 ath0s wrote:
The reason the overlap is so significant is because people are generally playing those with similar skill levels. Just imagine an outlier bronze with 80SQ who focuses heavily on spending money playing against an outlier GM with 80SQ who wins with strategy/multiprong micro (and hence doesnt spend as well as other GM)
Guaranteed in this matchup the GM would still sit pretty with 80SQ but the bronze macro will be taxed and he wont be spending his money nearly as well. Hence why your SQ doesn't determine your rating, its correlation, not a proof.
Edit: but still a cool correlation! Good point indeed, thanks. That's a similar issue to the ladder score - which is why people say "I'm <N> <League>", instead of just <N>. Similarly when talking about their SQ it helps if they say <SQ> <League> instead of just <SQ>.
My conclusions so far: use SQ to compare periods of development (larger samples), because it may vary a lot between individual games, and that won't mean your macro skill has improved. Don't compare it too much with SQ in leagues of very different order. Higher leagues have more compact SQ distributions, because they have smaller number of players, so the variations aren't as wide. As about maxed situations, as long as the map isn't starved, optimizing SQ stimulates you to build more static defenses, macro bases, production buildings for faster army remaking, and to attack/harass asap when you can. Overall, if used with care, it can be a very cool and useful tool.
|
The average over my last 15 games was something like ~65.
Though I noticed that the shorter games I had were upwards of ~90, but the longer they went the lower the SQ.
As sad as it is, being at least in the mid-Master range, I see how that happens, since I play a style that tends to stockpile large amounts of resources through mid-game to bludgeon the face of my opponent in with lots of tier 3 tech and remaxing in the late-game.
I think I'm going to start keeping a running tally of my SQ when I remember to, in order to get a long-term view of it. Maybe I'll do it monthly, even, in order to view my macro improvement.
|
Looks like I need to work some more on my decision-making, out of my last 25 games, those were 21 of them (removed 6pools I received ZvZ).
Note I'm diamond D:
+ Show Spoiler +
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/VnMlh.jpg)
