Rating Formula

Ratings are based on a specific formula that factors in your current rating, the rating of your competition, the number of rated games you've played and your win expectancy, which is calculated using the difference between your rating and your competitor's (or the average of your competitors' ratings if you are playing against two or more people). If you have not yet played any rated games, your current rating is set to 1200.
Ratings are determined based on a modification of the widely used system developed by Prof. Arpad Elo, former president of the American Chess Federation. Upon completion of a rated game, each player's new rating is calculated according to the following formula:

Rn=Ro+K(W-We)

where:
• Rn = new rating.
• Ro = rating before playing a game (new players start with a rating of 1200).
• K = 128 if player has played fewer than 10 rated games; or
o 64 if player has played 10 or more but fewer than 20 rated games; or
o 32 if player's rating is below 2100 after 20 or more rated games; or
o 24 if player's rating is 2100 or more but below 2400 after 20 or more rated games.
o If none of the above conditions is true, K = 16
• W = score for the game you just completed. A win is worth 1 point, a tie is worth 0.5 points and a loss is worth 0 points.
• We = expected score (win expectancy), as determined by the following formula:


We=one over 10(dr over 400)+1

where dr is the difference between the player's rating and his or her competitor's (or the average of the competitors' ratings if there are two or more in the game).

If there is a mistake anywhere i will edit it. If you think this rating formula is wrong let me know.

Originally posted by tlai1992
Ratings are based on a specific formula that factors in your current rating, the rating of your competition, the number of rated games you've played and your win expectancy, which is calculated using the difference between your rating and your competitor's (or the average of your competitors' ratings if you are playing against two or more people). If you ...[text shortened]... ere is a mistake anywhere i will edit it. If you think this rating formula is wrong let me know.

so if you are a 1200 played less than 10 games and you beat a 1300

1291=1200+128(1-.285714)

The formula you describe is not used until a player has completed 20 rated games. Up to that point, a provisional rating formula is used.

Provisional ratings are calculated as an average of the 'wild guess' ratings obtained from the completed games so far, where the 'wild guess' is
G = Ro+K(W-0.5)
K = 800 if opponent has a normal rating and K = 400 if opponent has a provisional rating.
Ro and W are as in the other formula.
For instance, beating an oppnent with a regular rating of 1350 gives a wild guess rating of 1750; losing to one with a provisional rating of 1200 gives a wild guess rating of 1000.

The 'normal formula' also has the feature that the value of K is halved if the opponent has a provisional rating.

In both formulae, the rating of an oppoonent who has completed less than five games is always taken as 1200.

All this is explained on the Help / FAQ page.

[Edit] One other point - the concept of multiple opponents does not apply to games played on this site.

Hello,

Last week i won a game against someone with +/- 1100 points.
My rating was 1363. Nevertheless i won that game i lost 3 points;
my rating is now 1360 points.
Is that possible?
Thanks to answer this question
kris

Originally posted by kriscuypers
Hello,

Last week i won a game against someone with +/- 1100 points.
My rating was 1363. Nevertheless i won that game i lost 3 points;
my rating is now 1360 points.
Is that possible?
Thanks to answer this question
kris

provisionals are erratic

Originally posted by kriscuypers
Hello,

Last week i won a game against someone with +/- 1100 points.
My rating was 1363. Nevertheless i won that game i lost 3 points;
my rating is now 1360 points.
Is that possible?
Thanks to answer this question
kris

Yes, if your provisional rating is 200 points or more higher than another provisional player, your rating will go down a little when you win (and a lot if you lose!)

The easiest way to work out what will happen to your provisional rating when you complete a game is as follows:
Subtract your rating from their rating (in your example, 1363 - 1112 = -251)
Add 200 if you win against someone with a provisional rating, 400 if they have a normal rating;
subtract the same amounts if you lose; leave the number alone for a draw (in your example, -263 + 200 = -51)
Divide the result by the number of games you will have completed, including this one (in your case, -51 / 19 = -2.68) Round the result to the nearest whole number and this is what happens to your rating.

In your case, this formula only applies one more time. New players have the added complication that all the "real" provisional rating is hidden until you have finished five games. Nonetheless, you can see how the formula behaves by looking at the first few bars of anyone's ratings graph who has finished between five and 300 games.

tlai1992 has PM'd me to ask for an example of the normal formula, so here goes.

In Game 481865 I am playing dean8. His rating is 1478, mine is 1537. This gives a 'win expectancy' for me of 1 / (10^((1478-1537)/400) + 1) = 1 / (10^(-59/400) + 1) = 1 / (10^(-0.1475) + 1) = 1 / (0.712 + 1) = 1 / 1.712 = 0.58.

That is, based on our ratings, I am about 58% likely to win.

If dean8 was to resign 😀 before our ratings change, my rating would go up by K * (W - WE) = 32 * (1.0 - 0.58) = 32 * 0.42 = 13 points, to 1550 and his would fall by the same amount to 1465.

The trickiest bit of the above is the win expectancy. Note that 10^(whatever) means "10 to the power (whatever)". Your calculator may not know how to do this 😛

Originally posted by RolandYoung
tlai1992 has PM'd me to ask for an example of the normal formula, so here goes.

In Game 481865 I am playing dean8. His rating is 1478, mine is 1537. This gives a 'win expectancy' for me of 1 / (10^((1478-1537)/400) + 1) = 1 / (10^(-59/400) + 1) = 1 / (10^(-0.1475) + 1) = 1 / (0.712 + 1) = 1 / 1.712 = 0.58.

That is, based on our ratings, I ...[text shortened]... ^(whatever) means "10 to the power (whatever)". Your calculator may not know how to do this 😛

get the advanced Texas Instruments >33 TI34 is the cheapest and best to do this you dont need 81 etc.

Half the fun is wondering how far I will rise if I beat that peron who is higher than me, or the fear of getting roasted by a 700 ranked player - working oput the formula is why we have a computer

Originally posted by invigorate
Half the fun is wondering how far I will rise if I beat that peron who is higher than me, or the fear of getting roasted by a 700 ranked player - working oput the formula is why we have a computer

You are right man!

But look at the bright side too! Imagine thrashing up IronMan31! 😲😵

Originally posted by Kapil Gain
You are right man!

But look at the bright side too! Imagine thrashing up IronMan31! 😲😵

I may be wrong here, but is the maximum points you can win / lose not 32? So if a 700 ranked player beats Ironman31, they will gain 32 points, and (s)he will lose 32?

Originally posted by Alcra
I may be wrong here, but is the maximum points you can win / lose not 32? So if a 700 ranked player beats Ironman31, they will gain 32 points, and (s)he will lose 32?

32 for 0-2099, 24 for 2100-2399, 16 for 2400 and above.

So if I read this right if a 1300 player beat Ironman they would gain 32 points and Ironman would lose 24 points (soon to become 16 points when he crosses 2400).

Originally posted by Alcra
I may be wrong here, but is the maximum points you can win / lose not 32? So if a 700 ranked player beats Ironman31, they will gain 32 points, and (s)he will lose 32?

Not if you are a provisionally rated player! 😀

Originally posted by Kapil Gain
Not if you are a provisionally rated player! 😀

Well thanks for digging up the thread to add your infinite wisdom to it. It was definately needed.

Originally posted by XanthosNZ
Well thanks for digging up the thread to add your infinite wisdom to it. It was definately needed.

definitely