EPT
January 9 rating list - Page 4
| Forum Index > SC2 General |
|
birchman
Sweden393 Posts
| ||
|
DusTerr
2520 Posts
| ||
|
TheBB
Switzerland5133 Posts
On January 11 2013 09:52 Hier wrote: How can they weigh more when their weight is determined by a change in certainty level of a player's rating? That's why I was asking how players' rating is changed upon a sudden win rate change given varying certainties based on varying uniform game frequencies? Also, I didn't really see anything on decay. At this point, all I can really do is refer you to the paper and hope you understand it. Games are not explicitly weighted anywhere. There is an iterative maximisation procedure which causes an implicit weighing. Still, it's a weighing that we can discuss in qualitative terms. On January 11 2013 16:15 Shinta) wrote: I don't agree with your judgements They're not my judgements. You're not reading my personal opinion. I don't tell this thing what to do.  On January 11 2013 16:52 JJH777 wrote: The Hall of Fame list is so bad... Bomber being number 1 by so much doesn't make any sense to me at all. He's never even won a GSL. The foreigners and Puma are strange as well. The hall of fame list is a ranking of players who were ever ahead of their peers, and by how much and how long. Remember that it doesn't take overall rating into account. The top rating today is almost 1000 points higher than it was when Morrow was number 1, and most likely many, many of the players on the list today are better now than he was then, but that doesn't matter (to this list, anyway) because Morrow was ahead of his peers and those other players aren't ahead of theirs. If you want an absolute ranking you probably want this, although what it's trying to do is essentially an impossible problem. (Rating systems can only compare players of the same time period, not historically. That's why I called the latter list only "highest ratings achieved", not "best players ever".) I also want to say I think many of you are confusing "winning tournaments" with "playing well". Yes they are related, strongly so in fact, but not the same thing. Two lost games at the wrong time can cause you to lose a whole GSL (or indeed, any other tournament—nothing special about the GSL here). It's possible to win a GSL by going only 20–15 and it's possible to lose it by going 17–4. The latter player would soar in the rankings, the first not so much. The list of players who are or were definitely good enough to win a GSL, but who haven't, is quite long indeed. Also remember that the GSL makes up only a fraction of the data. | ||
|
LeLfe
France3160 Posts
| ||
|
paralleluniverse
4065 Posts
On January 10 2013 19:59 TheBB wrote: This time I would like some feedback from you guys, on two points. First, the HotS release is coming up. What to do? I was originally planning to do nothing and just let the ratings modify themselves (which happens quite quickly). I don't think the games are that different. The other thing I could do is reset all ratings to 1000 and start from scratch. This would mean that the ratings would become quite unreliable for a fairly long period of time (several months) until they readjust. The third option is that I keep the rating, but I reset the uncertainty. This would be akin to marking the WoL ratings as "less trusted", putting a far heavier weight on new results in HotS. Do nothing. To reset the uncertainties all to 500 or whatever the initial value is, is to claim that you're suddenly just as sure of the skill of a player who played 1000 games as a player who has played 10 games. The goal is to get the best possible estimate of skill and uncertainty. And there is no conceivable reason why choosing an uncertainty of 500 for everyone would suddenly be a more accurate estimate than whatever it was before HotS, say 30. It's basically fudging numbers for no reason. Like how science teachers will take marks off if you write in a test that 4.4/2.9 = 1.5172413. The "correct" answer is 1.5 because they claim that the remaining digits is measurement error, despite the fact that there is absolutely no reason why fudging the 0.0172413 part to 0.0000000 would make the answer any more accurate. If you do anything at all, it should be no more intrusive than increasing everyone's uncertainty by, say, 10%. Doing more would be needlessly fudging numbers by replacing a good estimate with numbers pulled out of something. | ||
|
TheBB
Switzerland5133 Posts
| ||
|
Grovbolle
Denmark3805 Posts
| ||
|
Conti
Germany2516 Posts
| ||
|
revel8
United Kingdom3022 Posts
| ||
|
Grovbolle
Denmark3805 Posts
On January 12 2013 01:03 revel8 wrote: I am curious as to why some tournaments matches are included but some are ignored. I noticed that Stephano's win in MSI Pro Cup 13 has been ignored. Also you include some qualifiers like the IEM Katowice qualifiers but ignore others such as the MLG Spring Arena qualifiers or the NASL 3 qualifiers. Why is that? What criteria did you use to inclue MSI Pro Cup #4 but not MSI Pro Cup #13? Just an oversight? Mainly time and knowledge of the results. Links would help, the more games the better. | ||
|
Conti
Germany2516 Posts
On January 12 2013 01:03 revel8 wrote: I am curious as to why some tournaments matches are included but some are ignored. I noticed that Stephano's win in MSI Pro Cup 13 has been ignored. Also you include some qualifiers like the IEM Katowice qualifiers but ignore others such as the MLG Spring Arena qualifiers or the NASL 3 qualifiers. Why is that? What criteria did you use to inclue MSI Pro Cup #4 but not MSI Pro Cup #13? Just an oversight? If there's anything missing, it's best to provide links to the brackets (From Liquipedia, for instance). The exact dates when the games were played are also needed. | ||
|
ZeroCartin
Costa Rica2390 Posts
| ||
|
revel8
United Kingdom3022 Posts
On January 12 2013 01:25 Conti wrote: If there's anything missing, it's best to provide links to the brackets (From Liquipedia, for instance). The exact dates when the games were played are also needed. OK These are just the results I've noticed being missed for one player, Stephano. They include wins over Socke, Titan, Polt, Hyun, Genius, Supernova, Taeja, Hero, Sniper, Vampire and Finale. For completeness sake they should be included unless there is some criteria I am unaware of that excludes them from the reckoning. ---------------------------------------------- NASL 3 qualifiers played 02-09-2012 to 03-03-2012 http://wiki.teamliquid.net/starcraft2/North_American_Star_League_Season_3/Season_3_Qualifier ------------------------------------------------ GD Studio Arena 1 played 04-29-2012 http://wiki.teamliquid.net/starcraft2/The_GD_Studio_-_The_Arena/1 ---------------------------------------------- MLG Spring Arena Euro qualifiers played 05-12-2012 to 05-13-2012 http://wiki.teamliquid.net/starcraft2/2012_MLG_Pro_Circuit/Spring/Qualifier/Invite/Europe ---------------------------------------------- MSI Pro Cup #13 played 08-28-2012 to 08-30-2012 http://wiki.teamliquid.net/starcraft2/MSI_Pro_Cup_13 ------------------------------------------------- Also I am not sure if you include team leagues but Stephano has beaten the following players in IPTL since joining EG: Taeja, Hero, Vampire, Finale and Sniper and lost to Monster and Seed. http://wiki.teamliquid.net/starcraft2/IGN_Pro_Team_League_Season_1/Premier_Division | ||
|
shabby
Norway6402 Posts
| ||
|
TheBB
Switzerland5133 Posts
On January 12 2013 01:03 revel8 wrote: I am curious as to why some tournaments matches are included but some are ignored. I noticed that Stephano's win in MSI Pro Cup 13 has been ignored. Also you include some qualifiers like the IEM Katowice qualifiers but ignore others such as the MLG Spring Arena qualifiers or the NASL 3 qualifiers. Why is that? What criteria did you use to inclue MSI Pro Cup #4 but not MSI Pro Cup #13? Just an oversight? It's actually a pretty monumental task, populating a database with "all" pro SC2 games ever played. We're only a handful of people doing this on a voluntary basis. If you notice something missing, as Grovbolle and Conti said, and just post a link to the results here in this thread. If you want I can get you a submitter account so you can push stuff yourself, too. I think it's pretty quick and painless. Pretty sure that we have the most complete database available anywhere at the time being, but as you have pointed out yourself, it's hardly perfect yet. Edit: Thanks a lot!! | ||
|
Conti
Germany2516 Posts
On January 12 2013 02:05 revel8 wrote: ---------------------------------------------- NASL 3 qualifiers played 02-09-2012 to 03-03-2012 http://wiki.teamliquid.net/starcraft2/North_American_Star_League_Season_3/Season_3_Qualifier ------------------------------------------------ This is a good example of a problem I've come across here and there, and why some results aren't in the database yet: We need the exact dates for the games. When a game was played matters for the ranking, so a vague date for a few dozen games isn't really useful, unfortunately. For this one, I found the official thread, which luckily has exact dates included. But there's other tournaments and teamleagues that do not, especially older ones. Thanks for digging these up, by the way! | ||
|
29 fps
United States5724 Posts
| ||
|
revel8
United Kingdom3022 Posts
On January 12 2013 02:07 TheBB wrote: It's actually a pretty monumental task, populating a database with "all" pro SC2 games ever played. We're only a handful of people doing this on a voluntary basis. If you notice something missing, as Grovbolle and Conti said, and just post a link to the results here in this thread. If you want I can get you a submitter account so you can push stuff yourself, too. I think it's pretty quick and painless. Pretty sure that we have the most complete database available anywhere at the time being, but as you have pointed out yourself, it's hardly perfect yet. Edit: Thanks a lot!! I don't want to be critical of your efforts, because it is alot of work and having looked through Stephano's match records it does seem pretty comprehensive (bar those 2012 results I noted above). I don't want to give the impression that I am whining or attempting to discredit your work. The opposite is true, I am impressed and pleased that such a sterling effort has been undertaken. I was just curious about some results that I had noticed were absent. It is perfectly understandable that some results were simply overlooked. 2012 had a ton of tournaments pretty much constantly and the games did not really ever stop. It is my limited understanding of your ranking system that means players get points based on the current rank of players they beat, so missing games can potentially significantly change the accuracy for numerous players. So adding in those results I have spotted were missing should lead to more complete and accurate rankings. Thanks for your efforts for the scene. I do appreciate it. | ||
|
Grovbolle
Denmark3805 Posts
On January 12 2013 02:26 revel8 wrote: I don't want to be critical of your efforts, because it is alot of work and having looked through Stephano's match records it does seem pretty comprehensive (bar those 2012 results I noted above). I don't want to give the impression that I am whining or attempting to discredit your work. The opposite is true, I am impressed and pleased that such a sterling effort has been undertaken. I was just curious about some results that I had noticed were absent. It is perfectly understandable that some results were simply overlooked. 2012 had a ton of tournaments pretty much constantly and the games did not really ever stop. It is my limited understanding of your ranking system that means players get points based on the current rank of players they beat, so missing games can potentially significantly change the accuracy for numerous players. So adding in those results I have spotted were missing should lead to more complete and accurate rankings. Thanks for your efforts for the scene. I do appreciate it. I'll take a look at the IPTL you linked when I get the time for it  The older results people want us to add the harder it gets to check if the results are already added under either "unknown event" or something else. | ||
|
2muchSWAG
76 Posts
| ||
| ||
