Question about ratings:

So I played this guy some teaching games, he was around 1200 and me around 16 bills. So given a 400+ rating difference, what would be the expected ratio of wins V losses, my loss vs my wins? It took him 53 games to kill me in one. Is that about the expected win loss ratio of 400 point difference?

Originally posted by sonhouse
So I played this guy some teaching games, he was around 1200 and me around 16 bills. So given a 400+ rating difference, what would be the expected ratio of wins V losses, my loss vs my wins? It took him 53 games to kill me in one. Is that about the expected win loss ratio of 400 point difference?

If the rating difference is 400, you have a 92% chance of winning.

Originally posted by KnightStalker47
If the rating difference is 400, you have a 92% chance of winning.

So that would be 92 games out of 100, lose 8 so 50 games, lose 4. So I have 50 games and lost one. Does the math go beyond 400 point difference? It looks like my percentage is about 98%, what rating difference would that be?

Originally posted by sonhouse
So that would be 92 games out of 100, lose 8 so 50 games, lose 4. So I have 50 games and lost one. Does the math go beyond 400 point difference? It looks like my percentage is about 98%, what rating difference would that be?

How many games does it take for rating win ratios to be accurate?

I think with >=800 difference most rating systems class the game as unrated, as the win/loss %<1.

So, in other words, the stronger player is statistically likely to win 100 games in a row.

A good rule of thumb concerning getting a statistical measure is "five." You need at the very least five of something to have any kind of confidence about the statistical measure you are making. Clearly, with 53 games and one loss, you have a decisive advantage. To get a measure of just how great, you'd have to wait until you lost at least 5 times.

Another rule of thumb is that chess ratings are a poor measure of anything. Arpad Elo is quoted as having said, "The process of rating players can be compared to the measurement of the position of a cork bobbing up and down on the surface of agitated water with a yard stick tied to a rope which is swaying in the wind."

Originally posted by byedidia
A good rule of thumb concerning getting a statistical measure is "five." You need at the very least five of something to have any kind of confidence about the statistical measure you are making. Clearly, with 53 games and one loss, you have a decisive advantage. To get a measure of just how great, you'd have to wait until you lost at least 5 times.

An ...[text shortened]... e surface of agitated water with a yard stick tied to a rope which is swaying in the wind."

This is an excellent point, and very well expressed. I have a chess friend (OTB, in my home club) who has always been 100-150 points below me in rating, but he won all 4 of our first games, and I only recently reversed the trend with my first win against him.

At the very least, they do not account for style, and I can certainly vouch that my performance varies considerably based on my opponent's style.

Ratings are valuable in the aggregate, but they can sometimes be problematic in individual cases.

Originally posted by Paul Leggett
This is an excellent point, and very well expressed. I have a chess friend who has always been 100-150 points below me in rating, but he won all 4 of our first games, and I only recently reversed the trend with my first win against him.

At the very least, they do not account for style, ...

Style accounts for a lot, I agree. I have struggled to beat a particular 1400 player here at RHP yet walked all over another 1700 player. I seem to play better against stronger players (have drawn a few 2000+ players in my time); maybe it helps me to rise to the occasion.

Another factor, I think, is that below the master level, players are sometimes unevenly advanced. One 1800 player might be pretty solid at endgames but weak in openings, whereas a 1600 player who happened to catch him off guard in an opening variation might be able to end the game before the other player could bring his endgame technique to bear.

//ratings are valuable in the aggregate, but they can sometimes be problematic in individual cases.// Election predictions are famous for predicting how large blocks of voters will vote (Alaska or Texas), but not at all for how any individual will vote. That's statistics.