|
On July 02 2013 00:24 aldochillbro wrote:
I don't understand why anyone would include proleague in these statistics, it's just illogical. These stats either mean something or mean nothing. when you put a metagame and sniping based proleague in there it most definitely means nothing. Looking at the real statistics(posted in this thread), the game looks really balanced. it's weird to see protoss's actually keeping up
Yeah, because preparing for your opponent and metagaming his style never happens outside of Proleague.
|
On July 02 2013 00:24 Quateras wrote:Show nested quote +On July 02 2013 00:05 _Search_ wrote:
Wonder how much TvZ will change once Terrans figure out how to stop Roach/Bane all-ins... You can't stop a roach bane all in with playing greedy, whatever you do to be able to hold it makes your build non greedy anymore and thus you aren't that far ahead in the midgame. Its like saying : you cant wait for Zergs to figure out how to stop 11/11 rax when they would go 3 hatch before pool. (every player not named Happy)
Or, more likely, they'll refine a certain technique that totally nullifies the major timings.
|
On July 02 2013 01:26 _Search_ wrote:Show nested quote +On July 02 2013 00:24 Quateras wrote:On July 02 2013 00:05 _Search_ wrote:
Wonder how much TvZ will change once Terrans figure out how to stop Roach/Bane all-ins... You can't stop a roach bane all in with playing greedy, whatever you do to be able to hold it makes your build non greedy anymore and thus you aren't that far ahead in the midgame. Its like saying : you cant wait for Zergs to figure out how to stop 11/11 rax when they would go 3 hatch before pool. (every player not named Happy) Or, more likely, they'll refine a certain technique that totally nullifies the major timings.
Which will be spending money on units, rather then two engi-bays, 4 raxes and a third cc all on the back of 6 hellions.
|
United States5225 Posts
On July 01 2013 20:16 Clazziquai10 wrote:
HAH! Finally Terrans are no longer winning more than 50% of non-mirror matchups! In your face, hellbat haters! Looks like it won't be nerfed soon! XD
I don't really play SC2 much anymore, but I watch it a lot, and I really dislike Hellbats. Despite the fact they may not lead to greatly imbalanced statistics, they lead to stale gameplay, as the Infestor did.
So maybe Terran does need a buff somewhere if the Hellbat is nerfed, but Terran as a race shouldn't rely on Hellbat drops every game, that is bad.
|
United States5225 Posts
Interesting to note that Aligulac's June report is quite different:
http://aligulac.com/reports/
TvZ is 52.6% winrate for Terran
PvZ is 49.5% winrate for Protoss
PvT is 48.3% winrate for Protoss
|
I wonder how Shellshock feels about being called Shellbot.
We need a pie chart for this.
|
United States97276 Posts
On July 02 2013 01:46 BronzeKnee wrote:Interesting to note that Aligulac's June report is quite different: http://aligulac.com/reports/TvZ is 52.6% winrate for Terran PvZ is 49.5% winrate for Protoss PvT is 48.3% winrate for Protoss
should be because Aligulac includes every game and the guy who did the stats filtered out all the weekly cups and stuff.
For the OP stats
Same as always:
All Premier Tournaments
All Major Tournaments
All Direct Qualifiers to Premier Tournaments (this does not include Qualifiers for Qualifiers *only incl. later stages)
All Premier Teamleagues (GSTL, ATC and PL)
All Monthly Finals (Go4SC2 and Zotac Top 16, no weekly tournaments)
as well as filtering out some of the qualifier rounds in his Korean stats
OSL
GSL
PL
OSL Qualifiers (Ro8+)
Kespa MLG Qualifiers (Ro16+)
GSTL
HSC VII KR Qualifiers (Ro16+)
|
On July 02 2013 01:46 BronzeKnee wrote:Interesting to note that Aligulac's June report is quite different: http://aligulac.com/reports/TvZ is 52.6% winrate for Terran PvZ is 49.5% winrate for Protoss PvT is 48.3% winrate for Protoss
And more accurate because of much bigger sample size.
|
United States5225 Posts
More interesting is the fact that a guy named "ChaosTerran" came up with the statistics saying that Terran has a losing winrate in both mirrors... conflict of interests?
In the end, this comes down an argument of what games to include. ChaosTerran "filtered" out some games, while Aligulac "filtered" out others. Something tells me Aligulac is far more objective, but that is something subjective that everyone can decide for themselves.
From the Aligulac FAQ:
How do you decide which games to add?
This question doesn't have an easy answer. We mostly decide this on a case-by-case basis. Generally we will add a round from a tournament if that round contains a significant number of already rated players. (Usually higher than 25% or so.) For the large regular cups this usually means somewhere around top 16 to 8.
One common exception to this rule is large national tournaments, which when rated would create a «rating bubble». We try to avoid this as best we can, but it's a difficult thing to do. We are more lenient with tournaments where a significant proportion of the participants regularly compete internationally (Korea, Germany, Poland and Sweden), or if the tournament is significant in another way (such as TeSL).
If a tournament isn't in the database, it could either be because we felt it didn't cut it, or it could just be we have missed it or forgotten. This work is done on a voluntary basis, after all. You could try asking us about it, or submitting it yourself.
The best way for the up-and-coming player to get an Aligulac rating is probably to play lots of open tournaments and LANs, and keep going at it until you reach a round with a fair number of notables.
|
On July 02 2013 01:50 keglu wrote:Show nested quote +On July 02 2013 01:46 BronzeKnee wrote:Interesting to note that Aligulac's June report is quite different: http://aligulac.com/reports/TvZ is 52.6% winrate for Terran PvZ is 49.5% winrate for Protoss PvT is 48.3% winrate for Protoss And more accurate because of much bigger sample size.
more accurate, less relevant
|
United States97276 Posts
If you look at what TheBB said in this thread, though, the win rates are converging and we are at the most balanced state we've been in HoTS so far in terms of win rates.
+ Show Spoiler [Aligulac report] +
|
United States5225 Posts
What would really be a worthwhile exercise (and I'll probably contact the Aligulac guy about this later) would be to use the Aligulac rating of each player, and then compare winrates by race. Thus if a heavily favored Protoss player like Dear defeated QXC, it wouldn't necessarily show that Protoss was imbalanced. However if QXC (who is a big underdog) defeated Dear, it might indicate imbalance, especially if Terran underdogs were defeating Protoss favorites on a regular basis when they normally did not.
This would help control for skill and overtime, this could lead to the most accurate form of discerning balance.
Many of the complaints people had with "patch-Zergs" were that imbalance was allowing less skilled players to defeat higher skill players. I believe this is an issue now with Terran. Lots of Terrans are having success that never had much before due to Hellbats and Widow Mines (Drunkenboi, Bunny... ect) defeating heavily favored Koreans.
On July 02 2013 01:58 Shellshock1122 wrote:If you look at what TheBB said in this thread, though, the win rates are converging and we are at the most balanced state we've been in HoTS so far in terms of win rates. + Show Spoiler [Aligulac report] +
Important to note that this is still far more imbalanced than at many points in WOL. Sign of progress in the right direction certainly, but is not success yet, and a patch may be necessary for TvZ in specific.
|
On July 02 2013 02:00 BronzeKnee wrote:
What would really be a worthwhile exercise (and I'll probably contact the Aligulac guy about this later) would be to use the Aligulac rating of each player, and then compare winrates by race. Thus if a heavily favored Protoss player like Dear defeated QXC, it wouldn't necessarily show that Protoss was imbalanced. However if QXC (who is a big underdog) defeated Dear, it might indicate imbalance, especially if Terran underdogs were defeating Protoss favorites on a regular basis when they normally did not.
No need, I (much like the NSA) am omnipresent and will immediately come when you say my name.
Yeah, I've thought about this. It'd be cool to try.
|
On July 02 2013 01:52 BronzeKnee wrote:
More interesting is the fact that a guy named "ChaosTerran" came up with the statistics saying that Terran has a losing winrate in both mirrors... conflict of interests?
In the end, this comes down an argument of what games to include. ChaosTerran "filtered" out some games, while Aligulac "filtered" out others. Something tells me Aligulac is far more objective, but that is something subjective that everyone can decide for themselves.
From the Aligulac FAQ:
How do you decide which games to add?
This question doesn't have an easy answer. We mostly decide this on a case-by-case basis. Generally we will add a round from a tournament if that round contains a significant number of already rated players. (Usually higher than 25% or so.) For the large regular cups this usually means somewhere around top 16 to 8.
One common exception to this rule is large national tournaments, which when rated would create a «rating bubble». We try to avoid this as best we can, but it's a difficult thing to do. We are more lenient with tournaments where a significant proportion of the participants regularly compete internationally (Korea, Germany, Poland and Sweden), or if the tournament is significant in another way (such as TeSL).
If a tournament isn't in the database, it could either be because we felt it didn't cut it, or it could just be we have missed it or forgotten. This work is done on a voluntary basis, after all. You could try asking us about it, or submitting it yourself.
The best way for the up-and-coming player to get an Aligulac rating is probably to play lots of open tournaments and LANs, and keep going at it until you reach a round with a fair number of notables.
I think it might be this ChaosTerran:
http://www.teamliquid.net/forum/viewmessage.php?topic_id=319705¤tpage=148#2949
He was a bit of a legend in the "Where have all the Terrans Gone?" thread of the past, and its 7 replacement threads. He had a solid terran bias if I remember correctly.
|
The obvious answer is to filter out all amateur games and bottom of the food chain pro level games. Aligulac does a much worse job at that, simply because they include every game that could possibly be included. There is no filtering and if there is it is purely subjective and not based on set parameters.
It's important to note that a bigger sample size is sometimes detrimental to accuracy of your results. If you include samples that do not accurately reflect the data you are trying to compile, you are effectively wasting time because you are polluting your own data set with samples that of a different data set.
You don't look at BMW's to determine what the average acceleration rate of a Mercedes is. Likewise you don't look at amateur games to determine which race wins more games at the pro level.
Blackmailing the person responsible for these statistics is absolutely pathetic and you people should be ashamed for that, this is the 3rd time these win rates were released (april, may, now june) and the person is blackmailed now for the first time because the results do not match the expectations of the community, which is largely the result of communal reinforcement and the use of factoids which replaced actual facts a long time ago, especially on teamliquid.
|
Many of the complaints people had with "patch-Zergs" were that imbalance was allowing less skilled players to defeat higher skill players. I believe this is an issue now with Terran. Lots of Terrans are having success that never had much before due to Hellbats and Widow Mines (Drunkenboi, Bunny... ect) defeating heavily favored Koreans.
I went to check and in WCS America challenger league, Drunkenboi lost to State 2-1 in the first round. Then, in group stages, he lost to both Jaedong and Jim, while taking out Ian to stay in challenger league. No top koreans beaten here.
Bunny took out Finale for the WCS Europe qualifer with a 2-1, lost to ForGG twice in premier league. Lost to MMA in the bracket stage and beat Finale again with a 2-1 score in the groupstage to get back in premier league. At homestorycup he beat Violet 2-0 and later on lost to MC 2-0. Beating Finale 2-1 twice with korea>EU lag isn't a terrible big upset, imo.
I can't find any other up and coming terran players (Dayshi did all kill MVP in the ATC but even their coach said that the MVP players coudn't control their units due to weather conditions I think it was.
Neeb beat Huk 2-0 but lost to Puck 2-0 the next round.
Could you please show some of these huge upsets? Because the games I'm seeing aren't exactly like the era of zergs in WoL, if it was we would be seeing 4+ terrans in every RO8 and TvT finals everywhere.
|
United States5225 Posts
On July 02 2013 02:24 SlixSC wrote:
The obvious answer is to filter out all amateur games and bottom of the food chain pro level games. Aligulac does a much worse job at that, simply because they include every game that could possibly be included. There is no filtering and if there is it is purely subjective and not based on set parameters.
It's important to note that a bigger sample size is sometimes detrimental to accuracy of your results. If you include samples that do not accurately reflect the data you are trying to compile, you are effectively wasting time because you are polluting your own data set with samples that of a different data set.
You don't look at BMW's to determine what the average acceleration rate of a Mercedes is. Likewise you don't look at amateur games to determine which race wins more games at the pro level.
Blackmailing the person responsible for these statistics is absolutely pathetic and you people should be ashamed for that, this is the 3rd time these win rates were released (april, may, now june) and the person is blackmailed now for the first time because the results do not match the expectations of the community, which is largely the result of communal reinforcement and the use of factoids which replaced actual facts a long time ago, especially on teamliquid.
Slow down here buddy. The source of the statistics is incredibly important. If you go to the NRA's website for information on gun control you're going to get different information than the Mayors Against Illegal guns. The truth is probably somewhere in the middle, which is why people who are interested in the truth often look to bi-partisan government studies and the ATF.
So I didn't blackmail anyone. I noted that we have two different sets of statistics for the same thing that show different results. I then stated that there could be a bias, as ChaosTerran is a Terran player, and his results show a low winrate for Terrans. This means that ChaosTerran might have an agenda, and the way he filters games could reflect that.
Aligulac doesn't seem to have an agenda.
Finally, this has nothing to do with community expectations (unless you were saying that the expectation of the community is for results of the same statistic to be the same when looked at by two different people, as they should be).
|
On July 02 2013 02:24 SlixSC wrote:
The obvious answer is to filter out all amateur games and bottom of the food chain pro level games. Aligulac does a much worse job at that, simply because they include every game that could possibly be included. There is no filtering and if there is it is purely subjective and not based on set parameters.
It's important to note that a bigger sample size is sometimes detrimental to accuracy of your results. If you include samples that do not accurately reflect the data you are trying to compile, you are effectively wasting time because you are polluting your own data set with samples that of a different data set.
You don't look at BMW's to determine what the average acceleration rate of a Mercedes is. Likewise you don't look at amateur games to determine which race wins more games at the pro level.
Blackmailing the person responsible for these statistics is absolutely pathetic and you people should be ashamed for that, this is the 3rd time these win rates were released (april, may, now june) and the person is blackmailed now for the first time because the results do not match the expectations of the community, which is largely the result of communal reinforcement and the use of factoids which replaced actual facts a long time ago, especially on teamliquid.
Considering the source of the data is always important when determining how closely it might reflect reality. Unless the source is proven to have released unbiased statistics in the past, it is reasonable for anyone to look at them with a skeptical view. It is easy to run a large set of different numbers and then pick the one that match the outcome you want, rather than the once that are most accurate. We are not saying this "ChaosTerran" did that, but we are reasonable to check into see if he has any past bias.
Also, you are misusing the word "Blackmail".
|
On July 02 2013 02:24 SlixSC wrote:
The obvious answer is to filter out all amateur games and bottom of the food chain pro level games. Aligulac does a much worse job at that, simply because they include every game that could possibly be included. There is no filtering and if there is it is purely subjective and not based on set parameters.
It's important to note that a bigger sample size is sometimes detrimental to accuracy of your results. If you include samples that do not accurately reflect the data you are trying to compile, you are effectively wasting time because you are polluting your own data set with samples that of a different data set.
You don't look at BMW's to determine what the average acceleration rate of a Mercedes is. Likewise you don't look at amateur games to determine which race wins more games at the pro level.
Blackmailing the person responsible for these statistics is absolutely pathetic and you people should be ashamed for that, this is the 3rd time these win rates were released (april, may, now june) and the person is blackmailed now for the first time because the results do not match the expectations of the community, which is largely the result of communal reinforcement and the use of factoids which replaced actual facts a long time ago, especially on teamliquid.
Aligulac's purpose is not to collect game balance statistics, but to provide a reliable match prediction system. You are incorrect that there is no filtering involved, we tend to only add new players if they defeat someone already in the database, which would indicate potential of being a good player. Therefore the database consists of pros (I'm sorry if you feel that some of them are "bottom of the food chain", they're still signed by teams and play against other pros more or less regularly) and high level amateur/semi-pros, all of which you will find are Master league or higher.
Our goal is to try to quantify, with reasonable accuracy, how good players are, not only overall but also at a specific date, and compare them based on that. So for that, we need as many games as possible, including online cups and lesser known, scarcely-advertised tournaments. It's simply how it works, if you want your statistics to have any relevance at all, the sample size must be big enough.
You would be surprised how often your mid-high pros lose in online tournaments and qualifiers to "bottom of the food chain" pros or even completely unknown players. Therefore having as many relevant games as possible helps prevent situations such as TLPD's #1 player being San, or Tefel and KingKong being that high up the list. ^^
|
On July 02 2013 02:24 SlixSC wrote:
Aligulac does a much worse job at that, simply because they include every game that could possibly be included.
Haha, no. I have great volunteers, but that's pushing it.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/zuYeUXi.png)
