|
On July 29 2007 21:03 Last Romantic wrote:
plus, TLPD seems to have some discrepancy with kespa site. Kespa says Reach is like 51 or 52% win vs Z, whereas TLPD gives him an atrocious 45%
Actually Kespa's profile records are not in phase with their own game lists. I guess the former take special events into account, while the latter don't. (TLPD will, eventually, anyway.)
We'll get our own (automatic) ranking system sometime in the future. It needs to consider at least the following criterias:
- win % ;
- number of games played ;
- importance of games won/lost.
If anyone has any idea/suggestion regarding this, I'd be glad to hear them. 
|
On July 30 2007 01:19 LosingID8 wrote:
why has nada played so many televised matches compared to others that debuted around or before his time O_O
oh, and also am i correct in assuming that a lot of televised games haven't been added to the database? i recall NaDa getting an award a while back for his 300th win on MBC (in other words, excluding games played on OGN) or something like that
Again, the database doesn't include any of the special events yet (except one of two at most). This will hopefully change shortly.
|
awesome read i totally disagree with your "overall" ranking though...
|
Really good work.
didn't know saviors ZvP stats were that good.
Pusan as the best PvP player was kinda surprising aswell..
|
On July 30 2007 02:01 CustomXSpunjah wrote:
awesome read i totally disagree with your "overall" ranking though...
It's merely a statistical analysis ranking. You can't disagree with it. It's not the three best players, simply the 3 best winning %'s of players who have played more than 50 games. It's pure numbers.
|
Hong Kong20321 Posts
cool post dude
interesting statistics haha jaedong, sea, and ..whoever as 'top 3' O_O
|
This is such a useful thread due to the context you can see all the top player's repsective stats and whatnot in. What is most interesting is the quite dominating records Savior/Bisu have despite the level of competition in this age, and that mirror matchups are consistently the weakness of a lot of the top players. (although Protoss top players tend to be weakest in PvZ Statistically, exception Bisu I suppose :>)
Edit. Pop, maybe I've missed something completely (probably have), but has Kespa not actually included the results for the WWI? I didn't notice them, that's all and you mentioning TLPD eventually would made me think. Otherwise his stats should be 68/39 aka 107 aka 63.55% or so.
|
Wow, Sea has some killer stats.
But remember, you also have to take into account their opponents. If someone's been owning it up in RO16 or RO32 or whatever, and then getting 1-2ed or 0-2ed out of RO8, he faces lesser skilled opponents than, say Chalrenge did in the group of death, where he went totally bad ass and knocked Nada out, but had a losing ratio.
Well, my post doesn't make sense but it's 4 in the morning. Maybe someone else will explain better (or already has)
|
On July 30 2007 02:55 HaXxorIzed wrote:
Edit. Pop, maybe I've missed something completely (probably have), but has Kespa not actually included the results for the WWI? I didn't notice them, that's all and you mentioning TLPD eventually would made me think. Otherwise his stats should be 68/39 aka 107 aka 63.55% or so.
Nope, Kespa doesn't store WWI's results. Neither does TLPD (at least for now).
Not sure WWI should be included though. It's not really "professional games".
|
Wow nice information, thank you very much!
Jaedong's overall is good but he hasn't played as much games as other ones, we should wait for him to play at least up to 100 games.
|
On July 30 2007 03:17 PoP wrote:
Nope, Kespa doesn't store WWI's results. Neither does TLPD (at least for now).
Not sure WWI should be included though. It's not really "professional games".
Fair enough, the involvement of Nonkoreans would go a long way towards influencing Kespa in that respect. I will admit I was surprised that they chose to leave out the Korean vs Korean games though.
|
Good topic. Very informative.
|
since you corrected chrh's TvP, perhaps you should correct savior as well since one of his vZ was as a terran.
|
On July 30 2007 01:46 PoP wrote:Show nested quote +On July 29 2007 21:03 Last Romantic wrote:
plus, TLPD seems to have some discrepancy with kespa site. Kespa says Reach is like 51 or 52% win vs Z, whereas TLPD gives him an atrocious 45% Actually Kespa's profile records are not in phase with their own game lists. I guess the former take special events into account, while the latter don't. (TLPD will, eventually, anyway.) We'll get our own (automatic) ranking system sometime in the future. It needs to consider at least the following criterias: - win % ; - number of games played ; - importance of games won/lost. If anyone has any idea/suggestion regarding this, I'd be glad to hear them.
With the list of all the games and who won and lost you can calculate Elo ratings for each player (by calculating the change for each player after each game and having some starting value like 2000) and hence have a much more accurate/objective rating than any other method. Elo ratings are good because they accurately take into account the skill of a player's opponents.
|
United States20661 Posts
On July 30 2007 06:57 gravity wrote:Show nested quote +On July 30 2007 01:46 PoP wrote:On July 29 2007 21:03 Last Romantic wrote:
plus, TLPD seems to have some discrepancy with kespa site. Kespa says Reach is like 51 or 52% win vs Z, whereas TLPD gives him an atrocious 45% Actually Kespa's profile records are not in phase with their own game lists. I guess the former take special events into account, while the latter don't. (TLPD will, eventually, anyway.) We'll get our own (automatic) ranking system sometime in the future. It needs to consider at least the following criterias: - win % ; - number of games played ; - importance of games won/lost. If anyone has any idea/suggestion regarding this, I'd be glad to hear them. With the list of all the games and who won and lost you can calculate Elo ratings for each player (by calculating the change for each player after each game and having some starting value like 2000) and hence have a much more accurate/objective rating than any other method. Elo ratings are good because they accurately take into account the skill of a player's opponents.
while elo is really nice for considering % and relative skill of players, it fails to recognize importance of games; an OCL game would be weighted equally as an OSL game
|
I'm just glad to see GooDFrienD on every single one of the lists that includes T, even if he is at the bottom over all .
|
On July 30 2007 07:16 Last Romantic wrote:Show nested quote +On July 30 2007 06:57 gravity wrote:On July 30 2007 01:46 PoP wrote:On July 29 2007 21:03 Last Romantic wrote:
plus, TLPD seems to have some discrepancy with kespa site. Kespa says Reach is like 51 or 52% win vs Z, whereas TLPD gives him an atrocious 45% Actually Kespa's profile records are not in phase with their own game lists. I guess the former take special events into account, while the latter don't. (TLPD will, eventually, anyway.) We'll get our own (automatic) ranking system sometime in the future. It needs to consider at least the following criterias: - win % ; - number of games played ; - importance of games won/lost. If anyone has any idea/suggestion regarding this, I'd be glad to hear them. With the list of all the games and who won and lost you can calculate Elo ratings for each player (by calculating the change for each player after each game and having some starting value like 2000) and hence have a much more accurate/objective rating than any other method. Elo ratings are good because they accurately take into account the skill of a player's opponents. while elo is really nice for considering % and relative skill of players, it fails to recognize importance of games; an OCL game would be weighted equally as an OSL game
The importance of games is completely subjective so it can't/shouldn't be included in an objective rating system. Although it would be possible to assign a higher k value to "more important" games (so that ratings change more after "big" games) if you wanted.
|
I'm pretty clueless as to how the Elo ranking exactly works, but any system which takes the opponent's own ranking into account needs to be calculated in chronological order to work out, since the number of earned points directly depends on the rank of the opponent at that point, which in turns depends on the opponent's own games, etc etc.
Therefore we can't do this very accurately because we don't store the exact timestamp of the games yet. Not all bo5's are in order, and when two matches happened during the same day we don't know which one was played first.
Anyway, there are literally tons of way to build an automatized ranking system on TLPD, as well as many factors we could take into account (or choose to ignore). No idea which path we should follow nor how much weight we should give to each criteria.
|
On July 30 2007 08:06 PoP wrote:
I'm pretty clueless as to how the Elo ranking exactly works, but any system which takes the opponent's own ranking into account needs to be calculated in chronological order to work out, since the number of earned points directly depends on the rank of the opponent at that point, which in turns depends on the opponent's own games, etc etc.
Therefore we can't do this very accurately because we don't store the exact timestamp of the games yet. Not all bo5's are in order, and when two matches happened during the same day we don't know which one was played first.
For multiple games played in one day (or one match) you don't necessarily need to know the order since the typical way to do it would be to find the total result of the day/match (eg 3-2 would be +1 for the winner and -1 for the loser, 3-0 would be +3 for the winner and -3 for the loser) and use that rather than doing them one game at a time.
|
I updated sAviOr's stats to include the latest ODT games. I also took a loss off his ZvZ stats since one of his losses was when he was playing TvZ against GoRush. That actually moved him up 3 spots in ZvZ since everyone is so close in that ranking.
|
|
|
|
|
|
EPT
