- 27 Sep '06 20:49 / 1 editConfused.

I thought this was like OTB, after 400 points difference you get nothing.

Meaning my win should have got me 0 points.

Are you sure it's 600 for this site?

Mostly its 400 everywhere.

Edit: I check the calculations. It is 600. That's strange though.

The way I learn it was that if you play someone similar to you and win, you get 16 points. Every 100 point difference you get +/- 4 points.

So you beat someone 200 points higher and you get 24 points.

Maximum is +/- 32.

And those all based on 400 point difference maximum. - 28 Sep '06 00:04 / 1 editThread 50455

*Originally posted by wormwood*

Win Expectancy = 1 / (10^((OpponentRating-YourRating)/400)+1).

the 32 is the amount you win if you're*exactly*400 pts lower, or lose to someone 400 pts lower. if you win a 600pts higher opponent, you'll get more than 32.

but if you win a 400 pts lower rated player, it's only

32*( 1 - (1/(10^(-400/400)+1)) = 32*(1-(1/1.1) = 2.9 points.

and if I got it right, the (1-winexp) is always positive against lower rated but never zero, so in*theory*there's no limit for gaining rating against how ever weak opponents. but in practice the gain will be less than 1 point when you win against someone rated around 600 pts lower. and**if**there's internal rounding, there will be a limit.

okay, for the 'fun' of it, lets calculate when the gain is*exactly*1 point (for under 2100 rated, with k = 32 that is):

32*(1 - (1/(10^(d/400)+1))) = 1

1-(1/(10^(d/400)+1)) = 1/32

1/(10^(d/400)+1) = -(1/32 - 1)

10^(d/400) = -1/(1/32 - 1) - 1

d = 400*log10(0.032258) =**-596.54 points**difference wins you exactly 1 point.

or,

10^(d/400) = -1/(0.5/32 - 1) - 1

d =**-719.73**points difference, if the gain is rounded from 0.5 to 1 point.

or,

10^(d/400) = -1/(1.5/32 - 1) - 1

d =**-523,28**if it's floored from 1.5 to 1 point. - 28 Sep '06 00:16

I looked at those formula's and I ain't getting exactly the right answers.*Originally posted by wormwood***Thread 50455**if it's floored from 1.5 to 1 point.[/b]

*Originally posted by wormwood*

Win Expectancy = 1 / (10^((OpponentRating-YourRating)/400)+1).

the 32 is the amount you win if you're*exactly*400 pts lower, or lose to someone 400 pts lower. if you win a 600pts higher opponent, you'll get more than 32.

but if you win a 400 pts lower rated player, it's only ...[text shortened]... (d/400) = -1/(1.5/32 - 1) - 1

d = [b]-523,28

You are 1400 you play against a 1800 and win.

1800-1400 = 400/ 400 = 1

10 ^1 = 10

10+1 = 11 1/11 = 0.0909 repeating

1-0.0909...= 0.9090....

0.9090*32 = 29.0909

Not 32 points.

600 points higher and you win = 31 points.

Formula wrong, Calculator wrong or is it me? - 28 Sep '06 12:08

oh, you're right. winexp against 400 pts higher is 9.09%*Originally posted by RahimK***I looked at those formula's and I ain't getting exactly the right answers.**

You are 1400 you play against a 1800 and win.

1800-1400 = 400/ 400 = 1

10 ^1 = 10

10+1 = 11 1/11 = 0.0909 repeating

1-0.0909...= 0.9090....

0.9090*32 = 29.0909

Not 32 points.

600 points higher and you win = 31 points.

Formula wrong, Calculator wrong or is it me?

gain = 32*(1-0.0909...) = 29.09 just like you said. you also see here that you'd need an infinite rating difference to gain exactly 32 points.

so now I'm wondering where does the 400 in the original formula actually come from? is it an arbitrary statement that says "you will win 1 game and lose 10 with a rating difference of 400 pts"? I think it is. that statement is what scales the distribution of all ratings to its current shape, and the 32 is just a convergence factor which determines the step size (ie. speed) this convergence happens with. - 28 Sep '06 12:51

Performance can't be measured absolutely; it can only be inferred from wins and losses. Ratings therefore have meaning only relative to other ratings. Therefore, both the average and the spread of ratings can be arbitrarily chosen. Élő suggested scaling ratings so that a difference of 200 rating points in chess would mean that the stronger player has an expected score of approximately 0.75....*Originally posted by wormwood***oh, you're right. winexp against 400 pts higher is 9.09%**

gain = 32*(1-0.0909...) = 29.09 just like you said. you also see here that you'd need an infinite rating difference to gain exactly 32 points.

so now I'm wondering where does the 400 in the original formula actually come from? is it an arbitrary statement that says "you will win 1 game and los ...[text shortened]... onvergence factor which determines the step size (ie. speed) this convergence happens with. - 28 Sep '06 17:03

It was all a lie they told me when I joined the chess club. 32 points for a win over a 400 player. What a lie! you get 29.44 points for that .*Originally posted by wormwood***oh, you're right. winexp against 400 pts higher is 9.09%**

gain = 32*(1-0.0909...) = 29.09 just like you said. you also see here that you'd need an infinite rating difference to gain exactly 32 points.

so now I'm wondering where does the 400 in the original formula actually come from? is it an arbitrary statement that says "you will win 1 game and los ...[text shortened]... onvergence factor which determines the step size (ie. speed) this convergence happens with.

http://www.chess.ca/ratings_system.htm

Same thing just easier to use. - 29 Sep '06 20:21

Aren't you allowed to play friends :'(*Originally posted by z00t***RahimK, I thought you were more sensible than that! Shame on you. Are you among the players who target 1200-1600 players for easy pickings?**

The other guy I'm playing is a provisional brand new player. We used to play on Yahoo 3 years ago. Then he went to some US Navy Boot camp or something. Haven't talked to him for 3 years. The one days I get a PM and suprise. It's like I have a stalker! hehe...

He must have good searching skills to find me on Rhp.