|
Poland3751 Posts
I agree that in the long run ELO is more or less accurate. But what it may not reflect so well is short-term changes.
Or maybe in other words. ELO as a system was designed to punish looses as much as to award wins - I think you can agree on that. It's pretty symmetrical.
But if you look at the ELO changes for Brood Wars players it's easy to see that it's easier to suddenly improve your elo rating than to loose it. It may be only natural - I'm not sure. The idea is: while ELO is symmetrical on match-level it is not symmetrical on knockout tournament-level which seems not perfectly OK to me.
I remember that I liked the idea of Glicko system which introduced uncertainty based on the number of "missed" playing periods.
To clear it - I'm not saying the ELO is totally wrong and I enjoy comparing various ELO stats and treat them as good indicator. I just wonder if there isn't something better for Starcraft.
|
ELO greatly punishes losses for high rated players and greatly rewards wins for low rated players. So it's easier for *low rated* players to suddenly improve ELO, but the opposite is true for high rated players. I don't see the problem you seem to see.
|
I think some people are getting the wrong idea about ELO.
ELO is basically fair in terms of judging games by themselves. For example, Stork losing to Flash would reflect only a small drop in ELO, but losing to Cuteangel will result in a big one. Basically you match up the players and their respective strengths(ELO points) and judge based on that. Taken abstractly, it might just be the fairest way to judge players.
What elo doesn't take into consideration is when and what the games are for.
A good example would be Stork vs Hiya and then Modesty in the OSL. Those were Ro8 and Semi-final matches, incredibly important, and well prepared by all players involved. In fact, we can all see that Hiya played much better in the OSL than his previous losing streak suggest. And Modesty play inspired ZvP to get into the semi-finals. But Stork barely gained any ELO during this time. In fact, I'm pretty sure he lose ELO by beating Hiya 2-1(if it works the same as chess rankings).
The other problem with ELO is that there's no penalty for not playing games, and vice versa for playing more. Hence you have Skyhigh's 3rd best TvT ranking despite winning one TvT in the past half year. OTOH, you have TBLS basically playing gazillion games in the past month, and that's obviously quite important in terms of helping your team or your individual success.
Overall, ELO's a very good rating system as long as you understand the context and adjust to it.
|
It works the same as chess rankings.
As a matter of fact Arpad Elo was chess master.
|
On January 24 2011 23:44 mustaju wrote:Show nested quote +On January 24 2011 23:33 Holgerius wrote:On January 24 2011 23:29 mustaju wrote:On January 24 2011 23:25 ShadeR wrote:On January 24 2011 23:18 Lightwip wrote:On January 24 2011 23:12 POWEROUTAGE wrote:On January 24 2011 23:05 Lightwip wrote:
Yeah, he would rank pretty high, but probably under Jaedong right now because JD has a league on him. I doubt he would be considered for #2/3 if he didn't have those past achievements. I'm sorry if I sound condescending but you can't be serious. Jaedong's play has been terrible so far. Losing to violet in that manner, shaky play against woongjin, losing to stork... I think I've made my point. If anything, JD should drop in the PR. And if singlehandedly destroying SKT and the team that has all 3 of the other best players most well-equipped to take down Flash isn't enough for you to "consider" then you'll probably never accept that Flash is the better player even when he is. Oh no, I agree completely. JD has been having an awful month. But if only JD had the success in the previous months then he would be given more benefit of the doubt than Flash.. What if we consider that Jaedong is the strongest contender for the MSL championship? Flamewheel is fucked. You can't be serious. In his current form, he'll never get past Hydra. Dude, it's JD. I wouldn't use the word never if I were you. His name is Jaedong, but he currently plays like shit. That's how I'll choose to rate him.
Well for someone who's playing like shit its pretty crazy he's in the ro8 of the MSL and probably gonna 3-0 Snow. The only way I see Jaedong losing the MSL is to stork if stork loses to zero then I will be surprised if Jaedong doesn't win. If its Jaedong vs Stork will be an epic series though 
|
ELO glitches hard on BoX sets. It shouldn't be possible to lose points for winning a BoX.
That would be like losing points for winning a game by too small a margin.
|
I don't see why it is a problem to lose points by winning a BoX.
ELO is about adjusting the "skill" score of players based on how their results differ from the expected results. If player 1 is 200 points higher than player 2, winning only 2-1 means that the two players are actually closer in skill than their ELO points indicated, and thus need to be adjusted.
Also, it happens all the time that we lower our opinion of a player when they barely win against an opponent in professional sports, including starcraft. Example: see jaedong vs. Last from last night.
|
Canada2480 Posts
On January 26 2011 10:46 Severedevil wrote:
ELO glitches hard on BoX sets. It shouldn't be possible to lose points for winning a BoX.
That would be like losing points for winning a game by too small a margin.
if  Flash won a Bo3 2-1 against  ByuL, don't you agree he should lose some points?
|
On January 26 2011 12:18 nodule wrote:
I don't see why it is a problem to lose points by winning a BoX.
ELO is about adjusting the "skill" score of players based on how their results differ from the expected results. If player 1 is 200 points higher than player 2, winning only 2-1 means that the two players are actually closer in skill than their ELO points indicated, and thus need to be adjusted.
Also, it happens all the time that we lower our opinion of a player when they barely win against an opponent in professional sports, including starcraft. Example: see jaedong vs. Last from last night.
If you believe that SC is about how you play and not results, then that would be true.
Such a system would also mean that Flash won 40% gold and 60% silver in his loss against Effort in KA OSL. Or to take it even further, in Bisu's incredibly narrow 3-2 win against Stork in the MSL 3 years ago, do we say Bisu came out "slightly ahead" of Stork? Or that he won?
In the end, history is judged by winning and losing. And in BoX series, it's about winning or losing the BoX.
|
Stop trolling me; I never claimed that the historical importance of a series is numerically equivalent to how many games in the series were won or lost.
ELO is about measuring how likely one person is to defeat another and I stand by my statement that it isnt' a problem that it penalizes a player for losing games in a BoX they go on to win. I think that the Bisu-Stork series shows that the two players were close in skill, don't you think?
The bigger problem with ELO and series is that it varies depending on the order of the results. In the Flash-Effort series, flash would have ended up with a higher ELO if the games had gone LLWWF instead of WWLLL, which seems illogical.
|
I think the difference you're describing is quite insignificant, maybe 1-3 points worth of ELO change. Besides, I think of ELO as more accurate over a longer period of time.
|
On January 26 2011 15:37 nodule wrote:
Stop trolling me; I never claimed that the historical importance of a series is numerically equivalent to how many games in the series were won or lost.
ELO is about measuring how likely one person is to defeat another and I stand by my statement that it isnt' a problem that it penalizes a player for losing games in a BoX they go on to win. I think that the Bisu-Stork series shows that the two players were close in skill, don't you think?
The bigger problem with ELO and series is that it varies depending on the order of the results. In the Flash-Effort series, flash would have ended up with a higher ELO if the games had gone LLWWF instead of WWLLL, which seems illogical.
If that's the case, then we wouldn't have PL monsters like Sea or Leta or even pre-God-mode Flash. Nor would we have players like Kal or Stork who generally performs well in SL and loses a lot more than they should in PL. in Hence, the inability to differentiate between random PL matches and BoX series IS a problem when it comes to predicting one person beating another.
For example, Stork is Jaedong's bitch when it comes to SL play. Yet at the same times, Jaedong's Stork's bitch when it comes to PL play, with multiple Ace losses.
|
If you're just arguing that there is something to BW skill that ELO doesn't capture, of course I agree with that position.
It is a useful statistic that is interesting to think about, but it isn't everything.
|
My January Power Ranking at the moment - 26.01.2011.
1.  Stork 14-5
Wins: + Show Spoiler +Bisu, Hiya, Killer, Alone, Jaedong, Bisu, Jaehoon, Sea, Hyun, Tyson, Modesty, Modesty, Modesty, Hydra Hiya, Zero, Horang2, Zero, Zero
2. Bisu 13-3
Wins: + Show Spoiler +Calm, Hero, Kal, Bogus, Roro, Pure, Shine, Baby, Ggaemo, Firebathero, Iris, M18M, Zero
3. Jaedong 12-4
Wins: + Show Spoiler +Midas, Hero[Join], Midas, Baby, Midas, Free, Zero, Snow, Last, Bogus, Snow, Snow Stork, Violet, Snow, Snow
4. Flash 8-1
Wins: + Show Spoiler +Fantasy, Ssak, Bisu, Snow, Leta, Skyhigh, Bogus, Soulkey, Really
5. Hydra 10-5
Wins: + Show Spoiler +Classic, Roro, Classic, Action, Hoejja, Hero, Calm, JangBi, Brave, Great Midas, Stats, Bogus, Killer, Stork
6. Fantasy 9-6
Wins: + Show Spoiler +Stats, Action, Tempest, Calm, Calm, Calm, Sea, Soulkey, Free Flash, Calm, Calm, Jaehoon, Ggaemo, Zero
7. Zero 8-5
Wins: + Show Spoiler +Soulkey, Sea, Soulkey, Stork, Fantasy, Flash, Stork, Stork Soulkey, Frozen, Jaedong, Bisu, Action
8. Stats 7-4
Wins: + Show Spoiler +Best, Ruby, Peace, Ggaemo, Firebathero, Hydra, Hiya Fantasy, Snow, Bogus, Flying
9. Bogus 7-3
Wins: + Show Spoiler +Really, Hydra, Movie, Snow, Violet, Action, Stats
10. Tyson 7-2
Wins: + Show Spoiler +Really, Movie, Hogil, Firebathero, Ggaemo, Peace, Ruby
Runner-ups: Jaehoon (4-1); Sea (6-4); Turn (5-1); Violet (6-2)
|
wow, Bogus and Tyson are certainly not players we would expect to do well. What an interesting result.
|
ZerO's vT this month is so weird:
Wins against: Sea, Fantasy, Flash.
Loss against: Frozean.
Lol.
|
On January 27 2011 07:24 Goragoth wrote:
ZerO's vT this month is so weird:
Wins against: Sea, Fantasy, Flash.
Loss against: Frozean.
Lol.
...That's pretty much a summary of ZerO's career, honestly.
|
Estonia4504 Posts
On January 27 2011 07:34 Musoeun wrote:Show nested quote +On January 27 2011 07:24 Goragoth wrote:
ZerO's vT this month is so weird:
Wins against: Sea, Fantasy, Flash.
Loss against: Frozean.
Lol. ...That's pretty much a summary of ZerO's career, honestly.
And that's unfair to the game Frozean played. Drops friggin everywhere. ZerO didn't have a moment of rest.
|
On January 27 2011 07:34 Musoeun wrote:Show nested quote +On January 27 2011 07:24 Goragoth wrote:
ZerO's vT this month is so weird:
Wins against: Sea, Fantasy, Flash.
Loss against: Frozean.
Lol. ...That's pretty much a summary of ZerO's career, honestly.
You mean that sums up most of progamer's career. My point is that they try harder when playing against more famed opponent and win to get attention.
|
I'm not sure I like the placement of the people who have just blown through Ace. Sure, ggaemo is a legit win, but FBH, ruby, iris & co. aren't much opposition these days.
And don't forget Violet (6-2):
Wins: Jaedong, Killer, Perfectman, Kal, Hero, Calm
Losses: Hiya, Bogus
|
|
|
|
|
|
EPT
