|
On June 03 2013 06:00 Conti wrote:Hmm, are the 1200-rating-Koreans everyone with a Korean nationality in the database? Since then the system would favor players like SeoHyeon and KingKong, too, which is probably not the intention. An additional requirement could be that a player also had to have been in any of the Kespa/ESF teams at some point to slightly narrow down the list of naturally gifted players. (Hey, I might be a dev too, but I have nothing to do with the rating system, so I can ask stupid questions just like everyone else!  )
KingKong was part of Startale for a while. I think it'd be found that a high amount of low-profile Koreans have been part of an eSF team at some point in their career.
|
On June 03 2013 07:24 slowbacontron wrote:Show nested quote +On June 03 2013 06:00 Conti wrote:Hmm, are the 1200-rating-Koreans everyone with a Korean nationality in the database? Since then the system would favor players like SeoHyeon and KingKong, too, which is probably not the intention. An additional requirement could be that a player also had to have been in any of the Kespa/ESF teams at some point to slightly narrow down the list of naturally gifted players. (Hey, I might be a dev too, but I have nothing to do with the rating system, so I can ask stupid questions just like everyone else! ) KingKong was part of Startale for a while. I think it'd be found that a high amount of low-profile Koreans have been part of an eSF team at some point in their career.
Hah, true, I missed that. I was trying not to exclude all the high profile Koreans that left the Korean teams at some point. Eh, it'd still be slightly better. Not sure if it'd be worth the effort, though.
|
Mvp standing strong after beating up Europeans next to the Proleague monsters -_- Life due to relative inactivity?
oh, and it seems the records page has changed after the update to the algorithm, looks a bit more realistic. Mvp so high PP O_O
|
Have you guys done any cross referencing your winrates by map as well? Always curious to see likelihood/probability based off of the map pool.
|
On June 03 2013 08:24 FrodaN wrote:
Have you guys done any cross referencing your winrates by map as well? Always curious to see likelihood/probability based off of the map pool.
Our system completely abstracts maps out of the equation so to speak, and while I can't make a promise either way (and neither can others on the team, I would assume), there are no plans as of now to introduce them into the system. I will note however that we did not discuss this matter exhaustively.
My personal take on the idea of introducing maps into the mix, which is not representative of the team's opinion or anyone else's is as follows: it's simply not worth it. From a realistic, match submission standpoint, it would require far more effort to add maps to matches, even more so considering we already have almost 60k of them in the database already. Not only that, but some events we have extremely limited information on, which unfortunately doesn't include maps. So that's the practicality aspect. Assuming we had two dozen more dedicated volunteers than we already do, willing to scour the interwebs for details, wouldn't the payoff theoretically make this worth it? Well no. Ratings and predictions also being based off of maps might be more precise for people like Mvp, Stephano and MC who have played massive amounts of games, but for others it would only increase the uncertainty and volatility, simply because there aren't enough games played. This matter is exacerbated by the fact that map pools change, and most maps are neither Antiga nor Daybreak, in that they have much shorter a lifespan. In a world where there's 2 or 3 times as many tournament games played consistently, sure, it might well be worth it, but as it is I personally believe it would only add more uncertainty into the system.
Anyway I hope I didn't ramble too much and my response touched on what you were asking. If not, I at least hope it preemptively answered other questions that people might have raised.
^_^
|
On June 03 2013 08:51 MasterOfPuppets wrote:Show nested quote +On June 03 2013 08:24 FrodaN wrote:
Have you guys done any cross referencing your winrates by map as well? Always curious to see likelihood/probability based off of the map pool. Our system completely abstracts maps out of the equation so to speak, and while I can't make a promise either way (and neither can others on the team, I would assume), there are no plans as of now to introduce them into the system. I will note however that we did not discuss this matter exhaustively. My personal take on the idea of introducing maps into the mix, which is not representative of the team's opinion or anyone else's is as follows: it's simply not worth it. From a realistic, match submission standpoint, it would require far more effort to add maps to matches, even more so considering we already have almost 60k of them in the database already. Not only that, but some events we have extremely limited information on, which unfortunately doesn't include maps. So that's the practicality aspect. Assuming we had two dozen more dedicated volunteers than we already do, willing to scour the interwebs for details, wouldn't the payoff theoretically make this worth it? Well no. Ratings and predictions also being based off of maps might be more precise for people like Mvp, Stephano and MC who have played massive amounts of games, but for others it would only increase the uncertainty and volatility, simply because there aren't enough games played. This matter is exacerbated by the fact that map pools change, and most maps are neither Antiga nor Daybreak, in that they have much shorter a lifespan. In a world where there's 2 or 3 times as many tournament games played consistently, sure, it might well be worth it, but as it is I personally believe it would only add more uncertainty into the system. Anyway I hope I didn't ramble too much and my response touched on what you were asking. If not, I at least hope it preemptively answered other questions that people might have raised. ^_^
We have almost 60k matches, but more than twice the amount of games. So yeah, 130.000 games need to be classified if maps are to be included 
|
On June 03 2013 06:00 Conti wrote:
Hmm, are the 1200-rating-Koreans everyone with a Korean nationality in the database?
Yeah. We could use some traction on #46 before we get more advanced than that.
I still think it's "good enough". Players who are wildly over- or underrated compared to the pool they meet regularly (such as "foreigner" Koreans in this case) will readjust very quickly.
|
Suggestion for ranking pages (e.g. http://aligulac.com/periods/85):
If it doesn't slow page generation down too much, adding in coloured +/- values for rank and ratings instead of the arrows would remove the necessity of browsing through a lot of pages to find those numbers.
|
|
|
United States33578 Posts
time. based. decay.
#darkside
|
On June 13 2013 06:29 Waxangel wrote:
time. based. decay.
#darkside
*Hisssssssssssssssssssss*
|
suggestion threads best threads
|
On June 13 2013 06:29 Waxangel wrote:
time. based. decay.
#darkside
I don't think ratings decay can be incorporated into the glicko system. There is a linear time decay of deviation though it is currently unknown if it is significant by any measure. I'll will investigate that soon (this weekend probably) but since the linear coefficient is strongly linked to the predicting power I cannot guarantee that even if I find a way to improve it, we could implement it.
Still I do agree with you, we need to have a way to differentiate players who play consistently and thus deserve their ratings and those who don't.
|
|
|
|
|
Wow that's sick!
How did you get the rating numbers? The public dump doesn't have them, and I don't think anyone has computed them from source yet...?
|
I used the full database.
It's easy to download it from the site.
|
|
|
On June 15 2013 20:57 Arzar wrote:
I used the full database.
It's easy to download it from the site.
Technically it shouldn't be available to the public but the access isn't protected. I suppose you didn't compute anything then, just a simple extraction of the ratings.
|
It would be really cool if you could filter the match history depending on the opponents rating.
like the date option which is really awesome btw: (opponents rating between___ and____)
An other interesting stats besides the displayed winrate would be the average opponents rating.
Those two things would help to analyse the matchhistory tremendously.
one other thing i noticed. not sure if it is bugged but his vZ rating seems abnormal high: http://aligulac.com/players/286-CupCake/
<3 this site!
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/iUh31by.png)
