So when Dead Piggy among other people began to say that Heyheuhei was overrated, I decided I'd look into it to see how valid their claim is. So I took a gander through Heyheuhei's ladder record and was shocked. He has 92 total games counting towards his ladder rating at this very moment. Of these, he's lost 25. Say we round up and give him 8 more wins.. (since arguably he WAS at 100) This means he wins an even 75% of his ladder games. This is an accurate percentage as it is actually extrapolated from 100 total games. Has anyone done any better? Who's had the most games count toward their rating with the highest percentage of wins? Is it not very impressive that HHH can retain a top 15 ladder position over the course of 100 games?

I thought it was interesting and thought-provoking.

The rating is a more valid figure than win ratio, surely? We have the rating, I don't know what the point of consulting other statistics would be.

http://warlight.net/LadderTeam?LadderTeamID=277

X, the problem with rating is that for lower numbers of games (15ish), the rating can be off by as much as several hundred points because of the exact specifics of the games played. The larger the sample size, the more accurate the measure. What Lolowut is asking is if anyone else manages 75%+ win rate over a long period of games. Since you tend to play people around your own rating, that 75% will be a decent mix of players without picking on particularly bad players to inflate the number.

It's entirely possible that HHH has one of the few realistic ratings on the ladder, since he has a large enough sample of games to rule out standard variance.

as of this moment. played 144 (bit obsessive). lost 46. 68 % win rate. Gonna get better hopefully.

Since you tend to play people around your own rating

So, the win ratio is more valid than the rating because, in part, of the validity of the ratings?

I'm not saying the ratings are perfect - I don't have the knowledge to make any judgment on the rating system. I was just saying they're a more reliable metric than the win ratio. According to you, Heyheuhei's rating is especially accurate. So I think you actually agree with me that it is extraneous to dig up HHH's win ratio, when you already have the rating at hand.

Has anyone done any better?

Skunk and I are 100% on the 2v2 ladder.

X, Oliebol came up as 1st in the ladder with a 100% win ratio over 15 games with an incredibly high ladder rating. As he played more games, his rating went down, and as he continues I assume it will go down further.

Here's a question to ponder: How many games does it take for one to get to that 'realistic' rating value? 25? 50? 100? I bothered to include the win ratio, because it gives you a good idea of why he has a low rating value. (Meaning, the more games played, the more likely for his win ratio to go down, and the more likely the rating will go down)

The funny thing is, if you go ahead and figure out the win ratios of the top 10, you'll see something stunning. They all seem to go towards 75%. I am saying that, yeah, maybe HHH isn't playing as well lately, but to judge him on his rating makes little sense when he has played more games than anyone else on the ladder in the top 10. (Except Zibik)

Zibik has played somewhere around 75 games, and has won 74% of those. Is it entirely plausible that the longer the top players play in the ladder, the closer they get to this anomalous 75% win ratio?

*: When I say except Zibik, I meant that Zibik is the closest in the top 10 to HHH's amount of games, making his rating more valid when compared to HHH's.

Huruey: He has played 122 ladder games, which count towards his rating at the moment. He has lost 33. Which puts him at a 73% winratio.


I'm curious as to whether anyone has completed over 100 games counting toward their ladder rating and won above 80% of those games.

Let's say there is an perfect abstract performance rating at any time given for a specific player. That would be the expected value of the performance. The rating we see is the actual performance which, in mathematical terms, is a realization of a random variable. What we want from the ladder rating is for the realization of the variable to be close to expected value (meaning it measueres performance in a good way). The more games you play, the less variation we have. From experience it seems like 20-25 games is enough to give a rough accurate rating assuming the player is focused on the games. So basically in theory the more games you play, the more accurate the rating is. Unfortunately aside from that abstract there are important psychological factors. For example the more games you play or the faster you play the more you are prone to blunder. Also if you are ambitious and you want to reach a certain spot on the ladder it'll be hard for you with many unexpired losses holding you back. Also with playing 5 games at a time postponing your losses when you have 100 games already played won't have such an effect on your rating as it does when you have 15, that's why most newcomers get their highest rating quickly after joining the ladder. In general from a purely statistical point of view the more games you play, the more accurate rating you have. If you want to factor in psychological elements, you can, but I don't think it's really that important. The ladder is more about consistency than anything else, the rating will promote those who play carefully and focus 100% on each game, but that is pretty much how any other sport ratings work (yeah, I said it!).

X, Oliebol came up as 1st in the ladder with a 100% win ratio over 15 games with an incredibly high ladder rating. As he played more games, his rating went down, and as he continues I assume it will go down further.

So, again, the win ratio was less useful that the ratio in that case.

To reiterate, my point ultimately is that because the rating adjusts for the strength of opponents, it is more useful than a crude win ratio. Thus it is pointless to drudge up the ratio.

As szew says, the higher the sample size, the more accurate the statistic - this goes for the rating as much as it does the win ratio, so once again, the win ratio holds no advantage over the rating.

Am I not explaining myself clearly, or what? My point is self-evident, it shouldn't take 3+ posts to explain it.

So, again, the win ratio was less useful that the rating in that case.

>fix'd

I bothered to include the win ratio, because it gives you a good idea of why he has a low rating value. (Meaning, the more games played, the more likely for his win ratio to go down, and the more likely the rating will go down)

So you calculated his win ratio in order to prove that losing games makes your rating decrease? I don't understand. In the first post you asked for people who had a higher ratio, as if you would be surprised to find someone with one. Now you say he has a lower rating because it corresponds to the (lower) win ratio.

Ugh, I don't know why I bothered to break my forum silence over this. I hate these ladder ratings threads.

And to reiterate, X, I'm mainly curious as to who has pulled the highest percentage of wins (or highest rating, though it's harder to judge when some players are no longer player in the ladder, which is why I mention winrate. As I'm not familiar with many people not currently in ladder I'm pulling these up.) with the highest amount of games played.

or highest rating, though it's harder to judge when some players are no longer player in the ladder, which is why I mention winrate.

Well that makes sense, why the fuck didn't you say that instead of going on about Oliebol?

Incidentally, the sentence structure of that last post made my brain cry.

Am i right in thinking you are limited to playing only players ranked within 20 of your ranking? and that the matchmaking system is aiming for everyone to have a win ratio of 50%?

If that is the case, in the long run, anyone outside the top 20 (who play slightly more players below themselves leading to a higher win %) or bottom 20 (the opposite) should have a win ratio of 50%.

You also have to factor in the fact you also play players who are coming into ladder, who get inflated ratings off of 1-2 wins.

Yeah, Sir Bombs, but the ladder isn't so stable due to newcomers and retirees.

Three assumptions I have about the ladder:

1. The top active players: HHH, Oliebol, Zibik, Sze, Tim, dead pig,, et al.
2. When the top players have games against each other, at least half of the games are decided by luck (and not skill or risk management).
3. It is perfectly normal for any of the top players to get lucky in a series of 10 games against other top players. But across 20+ such games the winning percentage should tend towards 40-60%.
4. Psychology: The template is dull, and each person has his own tipping point. For me, having completed about 15-20 ladder games is when I start losing interest. If you play games after ladder burn out has set in, you focus and care less and lose more frequently.
5. I think playing at least 20 games against other top players on the ladder will lead to the most accurate statistics. But since that will include expired games, winning percentage (overall, vs #1s, vs #2s) should be used for comparisons.
6. Some players enter the ladder at an elite level (Sze, unknown, Zibik, dreuj, Mian, et al), others use the ladder as a means to reach an elite level (tim, myself, ace, chaos, et al). To compare only games played after reaching an elite level would be the most accurate way to compare the top players on a historical basis. I can safely say that about 60 ladder games ago I suddenly understood the game at an elite level. Before then I didn't quite understand picking and a few situational strategies peculiar to the map and settings.
7. Not all wins are wins and not all losses are losses. Sometimes people just quit in games when winning or before a winner has been clearly determined. Other times people can't make moves in time and are booted. To count all games without first sifting through the games that shouldn't be counted is not fair. Chris and zaeban, for example, were booted from at least 10 games each. Only statistically significant games (winner based on turns played) should be compared.

Sze, you could easily test all of our ladder theories. Increase your game count so you play more games against top players and approach 20 total games against top players. Then we can compare the totality of your games to see the trends, luck and skill.

**the not three ... fat fingers and autoselect