|
Contents
What is this?
Aligulac.com is an ongoing statistical project and website in development started in December 2012. It offers a comprehensive database of games from the pro and semipro SC2 scene, as well as a unique rating system aimed at rating players and teams, and predicting games. Every two weeks, when a rating list is finalized, I write a post like this.
The FAQ might be able to answer your questions. If not, I'll be keeping an eye on this thread so you can ask away here.
Also, before we start, I want to quote Heartland from one of my previous threads who had this to share, better put into words than I ever could.
On February 03 2013 02:50 Heartland wrote:
I think what's cool and great about this work is that it does what statistics are good for. They give you the ability to create data and then to look at it critically. But maybe people in this thread confuse statistics with the Truth with a capital p (sic). That's not the way you should read statistics, whether in the morning paper or on TL. Rather statistics can make us think about deeper connections that we haven't seen before, twist and turn around concepts and play with them through statistical models. They're not meant to say "Scarlett should be in Code S." Obviously there are flaws or issues with these stats, but it's common for statistics everywhere. What you can do with that is to add or change some modifier, let it meet other forms of reasoning or to extrapolate on what we take for granted.
So yeah, tl;dr, lies, damn lies and statistics are the case with all stats but it's not the point of stats.
So, let's get on with the news.
Twitter
I don't quite get the fascination with Twitter, but since it's one of those important e-sports things there is now an official Aligulac twitter @Sc2Aligulac. It's operated by my PR advisor Grovbolle, though I'll use it myself to announce features. You can follow this for predictions, announcements and the regular goofy stuff that Grovbolle will write. 
New features since last time- I am still tweaking the rating system, and you shouldn't expect it to settle just yet. It's a slow process, since whenever I change something I have to figure out the optimal parameters all over again, which is a very lengthy process. The latest change allows players who play rarely to adjust faster and players who play more often to adjust slower. Since most players are of the second type, I had to overall raise the speed of adjustments to compensate.
- There are now fewer than 10,000 uncatalogued matches, down from 13,000 last time. Almost all of this comes from Go4SC2 cups. (We're chasing that number zero.) Also, the number of games in the database passed 120,000.
- The results lists now have icons that indicate game version (WoL/HotS) and tournament type (offline/online).
- Player, team and event pages contain more information now. Events have matchup statistics, homepages, liquipedia links. Players have earnings and a ratio of offline matches. Same with teams. One interesting thing is the MC number. To get an MC number of 1, you have to have played an offline match against MC. To get an MC number of 2, you have to have played an offline match against someone with an MC number of 1, and so on. MC was chosen because he has more offline games than anyone else. Think of this as our version of Erdös or Bacon numbers. Basically, more numbers! Yay!
- There is a new balance report page. This will be expanded greatly as I think of more statistics to produce.
- Player match histories are now filterable by race, nationality, match length and type (offline/online). If you want to check the lifetime head-to-head record, the best way to do that is like this.
- The code has been put on GitHub. Along with the database dump this should allow you to replicate everything.
- Feature requests and bugs are best submitted here, though it's not that important.
- Charts have been improved. We ditched the server-rendered plots and went with javascript instead, which lightens server load and allows you to interact much more easily. Try it out. You can click and drag to zoom and export graphs.
Old features and things you should still know about
Everyone can submit results. Just go here: http://aligulac.com/add/ – some pro players have been scouring their histories and submitting the missing bits (mostly the wins cough, cough), which is great!
Help is still needed! The database is probably the biggest in the world of its kind and it's being maintained by a handful of people. There are still 10,000 matches that have not been catalogued and I dare not imagine how many games we have yet to add. Almost all the Playhem cups are still missing, for example. People volunteer fairly regularly, but the workload is too damn high for the current staff. For all the rejected TLPD volunteers (I know you are out there, I was one myself) this would be the perfect project.
Anyway, let's get to the juicy bit. These two weeks include games from (in no particular order): MLG Showdowns, Day[9] King of the Beta, IEM WC, GSL, Proleague, Ritmix RSL, Flamingcow.tv open, IPL qualifiers, Cascade1xbet league, IPTL, Francophone Championship, Gladiator's Arena, GSTL preseason, Assembly, Yegalisk Master Cup, aAa Pro Challenge, ESET UK Masters, and a bunch of minor regional leagues (Danish and Dutch mostly), showmatches and weekly cups.
Top 10 players
-
 Life 1887
-
 Bomber 1825
-
 PartinG 1815
- Leenock 1810
- DongRaeGu 1798
- Polt 1789
- HyuN 1777
- viOLet 1768
- ForGG 1766
- Rain 1712
The biggest winners are Polt (+65) and Violet (+41), and the biggest losers are DongRaeGu (-52) and Hyun (-41). Rain and ForGG also drop some points.
Top 10 foreigners
- VortiX 1708
- Stephano 1690
- Sen 1667
- LucifroN 1630
- Scarlett 1611
- Nerchio 1603
- Fraer 1553
- Snute 1545
- XiGua 1511
- LiveZerg 1493
The reason tournaments like IEM WC are so important is that it provides a «trading ground» for points to pass between foreigners and Koreans (usually from foreigners to Koreans). This time we see Vortix (-47), Lucifron (-71) Snute (-71) lose a lot of points. Nerchio and Fraer both gained some 20-odd points. They didn't take them from Koreans though...
Top 10 teams, all-kill edition
- Incredible Miracle 92.61%
- StarTale 92.46%
- MVP 90.99%
- AZUBU 90.40%
- SK Telecom T1 89.44%
- FXOpen e-Sports Korea 85.71%
- Prime 85.09%
- Team Liquid 83.73%
- Evil Geniuses 82.33%
- Team Acer 82.10%
Top 10 teams, proleague-style edition
- Incredible Miracle 80.51%
- AZUBU 77.22%
- MVP 76.54%
- StarTale 76.44%
- SK Telecom T1 69.46%
- Team Liquid 60.61%
- Prime 59.84%
- FXOpen e-Sports Korea 59.74%
- Evil Geniuses 58.85%
- Woongjin Stars 54.12%
Note that team ranks are not influenced by team tournaments, but by the individual results of the players themselves. That is why Woongjin, while dominating PL, are climbing so slowly. PL doesn't offer enough games to let their players climb fast enough, and this system is incapable of taking coach strategy into account.
Contact us
For comments, feedback, feature requests and especially volunteers, you can reach us by:
- This thread, duh.
- Twitter @Sc2Aligulac.
- PM. Me.
- IRC: #aligulac on quakenet.
- E-mail to evfonn(at)gmail(dot)com.
- Issue list on GitHub.
- Pilgrimage to Zürich. I will find you.
Thanks to- New guy: FrozENDruiD, manager of NewRoSoft.
- Programming team: Conti, Otolia (still getting ready
 ).
- Database maintainers: Conti, kiekaboe, Inflicted, Grovbolle, PhoenixVoid, scisyhp, Susurrus.
- PR manager and tweeter: Grovbolle
- Everyone who PM's and e-mails me with suggestions. You're all heard.
See you in two weeks!
|
o.O IM #1, nice, although I didn't expect that.
|
Oh, I forgot to mention. I'm in need of someone who can do some graphics. What I need is not much: 16x16-ish PNG graphics for WoL/HotS and online/offline icons that are easily distinguishable and intuitively clear. I don't think the current ones are up to snuff. Any takers?
|
On March 07 2013 19:58 EtherealDeath wrote:
o.O IM #1, nice, although I didn't expect that.
It's based on players overall skill level, and not on actual results in teamleagues. (Since IM has a lot of killer players)
Explained in a blog post:
The all-kill score is generated assuming a best-of-9 format, that is, the first team to eliminate five players on the opposing team wins (this is used in the GSTL, for example). The teams are assumed to always field their five best players (by rating) in increasing order. (This is not very accurate, but what can you do?) I have also allowed teams with fewer than five players to get an all-kill score. (Which will be very poor, of course. For SK to win an all-kill match, MC would have to win all the games himself, for example.)
For Proleague score, I am using a best-of-7 format (as seen in Proleague). The teams are assumed to field their best six players in a random order (again, not very realistic), and their best player in an eventual ace match. Teams with fewer than six ranked players won't get a proleague score.
|
How are you going to handle WoL vs HotS ranks? I honestly would like to see players ranked by their HotS performance and their HotS performance only, starting from 0 ( or whatever the starting point is ) instead of being based on their WoL score.
I would like to see WoL only rank, HotS only rank and WoL+HotS rank.
I dont know how hard it would be to implement, but it would be very interesting to be able to filter the ranks by a specific time frame and/or a specific tournament.
For example.
Get all the HotS events that happened before March 12th ( IEM, MLG Showdowns, GDstudio KOTH, GSTL HOTS and MLG Dallas ), I am counting MLG Dallas because that event is going to be weird since pros do not have access to HotS to practice pre event, and the showdown matches were played prior the release.
Then you do the wizardry math stuff and give me the best pre-hots release players, then I can compare that to all the post-release data once it becomes available.
For tournament specific, lets say GSL Season 2, it would be interesting later on to see during that specific tournament who is the the best player overall in that specific tournament vs the player that had the bigger growth. Some underdog that didnt make to the GSL Ro8 for example, but that came all the way from Code B with his 1600 rating and beat a lot of amazing players on his way up. You could have tournament specific ladders for the major ones like GSL, MLG, IEM, Dreamhack, so you are only compared with other people who played in that tournament, not the entire world.
I dont know if I explained well what I want, or if its even possible/worth your time to implement, but I felt like mentioning it anyway.
|
opterown
Australia54784 Posts
ah haha that zvt winrate after patch ;D
|
Unexpected
|
I like your balance report graphic; is there a way to see foreigners vs koreans over time in mirror matchups and balance by korea only?
|
Hey, fraer at least 3-0ed SortOf, that's not too bad.
|
On March 07 2013 20:30 Hiwashi wrote:
How are you going to handle WoL vs HotS ranks? I honestly would like to see players ranked by their HotS performance and their HotS performance only, starting from 0 ( or whatever the starting point is ) instead of being based on their WoL score.
I would like to see WoL only rank, HotS only rank and WoL+HotS rank.
It's something we have been talking about. As of right now, there is only one ranking though we can easily compile a WoL only ranking, it is not currently possible for HotS.
The issues that have been brought up are mostly in TheBB's garden so I'll let him explain that in details but a summary would be : how to calculate the HotS ranking ?
|
United Kingdom14103 Posts
Cool to see xigua top ten foreigners.
I always enjoy reading these, thanks!
|
It would be cool if you look at the difference between projected team results and actual team results over the course of team leagues to develop an effective value of the manager. (essentially determine how much a given coach improves the team over the team members' individual abilities.) and you could actually give coaches values and rankings and then incorporate them into the team predictions.
|
On March 08 2013 02:22 Tixian wrote:
It would be cool if you look at the difference between projected team results and actual team results over the course of team leagues to develop an effective value of the manager. (essentially determine how much a given coach improves the team over the team members' individual abilities.) and you could actually give coaches values and rankings and then incorporate them into the team predictions.
Small sample size would be a big issue for this. Currently everyone on the team agrees that the team all-kill/proleague scores are more for show than anything since they aren't that useful. That comes with the limitation of all the assumptions behind the calculation which I quoted in an earlier post.
|
ForGG seems overrated due to not playing top Koreans. The foreigners other than Stephano and Naniwa also seem overrated.
|
On March 08 2013 03:56 Salient wrote:
FORgg seems overrated due to not playing top Koreans. The foreigners other than Stephano and Naniwa also seem overrated.
Most likely, numbers can only take you so far, the more interscene matches the better for all.
|
What's this mysterious MC number?
|
On March 08 2013 04:20 MCXD wrote:
What's this mysterious MC number?
It works like this:- MC has an MC number of zero.
- Anyone who plays MC gets an MC number of one.
- Anyone who has not played MC but who has played someone who has played him gets an MC number of two.
- And so on. If you have played someone with an MC number of k but not someone with MC number of k–1 or lower, you get an MC number of k+1.
Compare with the Erdős number.
|
On March 08 2013 04:24 TheBB wrote:It works like this: - MC has an MC number of zero.
- Anyone who plays MC gets an MC number of one.
- Anyone who has not played MC but who has played someone who has played him gets an MC number of two.
- And so on. If you have played someone with an MC number of k but not someone with MC number of k–1 or lower, you get an MC number of k+1.
Compare with the Erdős number.
So it's a math meme and sort of homage to MC. I guess you picked MC due to his many high level games and record prize winnings. Out of curiosity, did you consider Mvp or Nestea as alternatives?
|
As always, your work is greatly appreciated.
Any update on how you plan to handle the HotS transition? Is a wipe of any kind planned, or do you let the system handle this by itself?
|
On March 08 2013 04:44 Salient wrote:Show nested quote +On March 08 2013 04:24 TheBB wrote:On March 08 2013 04:20 MCXD wrote:
What's this mysterious MC number? It works like this: - MC has an MC number of zero.
- Anyone who plays MC gets an MC number of one.
- Anyone who has not played MC but who has played someone who has played him gets an MC number of two.
- And so on. If you have played someone with an MC number of k but not someone with MC number of k–1 or lower, you get an MC number of k+1.
Compare with the Erdős number. So it's a math meme and sort of homage to MC. I guess you picked MC due to his many high level games and record prize winnings. Out of curiosity, did you consider Mvp or Nestea as alternatives?
MC has the most offline games, and has played lots of different gamers. It was goody before and not restricted to offline, resulting in everyone having a number of 1 or 2. (Goody has a shit ton of matches in the DB)
|
On March 07 2013 20:30 Hiwashi wrote:
How are you going to handle WoL vs HotS ranks? I honestly would like to see players ranked by their HotS performance and their HotS performance only, starting from 0 ( or whatever the starting point is ) instead of being based on their WoL score.
I would like to see WoL only rank, HotS only rank and WoL+HotS rank.
I dont know how hard it would be to implement, but it would be very interesting to be able to filter the ranks by a specific time frame and/or a specific tournament.
On March 08 2013 04:47 ACrow wrote:
As always, your work is greatly appreciated.
Any update on how you plan to handle the HotS transition? Is a wipe of any kind planned, or do you let the system handle this by itself?
Well, I had some ideas on this and I polled the audience a while back.
First, we could just wipe all the ratings and start from scratch. People would have to live with useless ratings for about six months (very rough estimate). You can check out the early ratings from WoL to see how long it took them to stabilize into something that seemed sensible.
Second, we could wipe the uncertainties and not the ratings. This would let all the ratings adjust faster but you're still producing crud for quite a while.
The problem with both of these approaches is that there is a period of overlap of about three-four months with a sparse amount of HotS games that we either have to ignore completely, artificially shift forward to the actual HotS launch or just «pretend» that they're WoL games.
This can be solved by producing two ratings, like Hiwashi suggests. However, this still raises the question of how to initialize the HotS series and when this is done. At a fixed time for each player or just whenever they play their first match? What exactly do they inherit from the WoL series? So basically you still have to answer all the above questions.
The solution with the least amount of question marks is just to roll with it and let the ratings adjust naturally. The question you should ask yourselves is this: which strategy will take the shortest time to stabilize into sensible HotS ratings? To provide an example, look at the current rating list. Then compare it with a totally wiped list where every player is rated 1000. Which of them do you think is closer to a sensible HotS rating right now? If you think it's the former then it makes no sense to totally wipe the list because you'll probably be in the «Dark Ages» for longer. I'm fairly confident (but not 100% sure of course) that the ratings will become sensible faster if we just leave them alone. The people who answered the poll tended to agree.
It's not difficult to publish two or three concurrent lists, but I question whether the increased complexity is really beneficial.
Still, I understand the concern. The HotS release should be noted... somehow. We're also planning on doing some numbers relating to patches, so this is is related. We'll see.
Input is welcome.
|
I would suggest waiting to make HoTS rankings until you have an arbitrary "large number" of games from pro players. Maybe wait for 3 months before doing the calculations. I would exclude pre-release beta games.
|
|
|
Thanks for the correction Tempest. All have been added.
|
Many thanks for the input, that's amazing. :D
|
|
|
|
|
|
|
|
EPT
