Computer Chess Club Archives


Search

Terms

Messages

Subject: USCF Rating Changes--Linear approximation

Author: Stephen A. Boak

Date: 13:55:20 08/26/00

Go up one level in this thread


On August 26, 2000 at 15:52:39, Uri Blass wrote:

>On August 26, 2000 at 14:46:08, Stephen A. Boak wrote:
>
>>In below threads, there is discussion about comparing performances of various
>>programs using 1) Ranking or average Ranking (for 'teams') or 2) TPR -
>>Tournament Performance Rating or average TPR (for 'teams').
>>
>>I suggest that using 2) TPR as a measure of which program (or team) is best is
>>inappropriate under many circumstances, including the circumstances of the just
>>concluded WMCC competition where the  ratings of the participants varied
>>greatly.
>>
>>Examples:
>>
>>My rating is approx 1900.  Let's say I play in a 4-round tournament in an
>>appropriate class section (Under 2000 rating--which typically includes mostly
>>1800 to 1999 rated players) that happens to several very weak entrants--possibly
>>up and coming young players that wish to get tougher competition to foster their
>>chess development.
>>
>>Example 1:
>>
>>Assume I play the following opponents, and have the following results:
>> Ro     Pts     GPR (Game Perf Rtg--using +/- 400 rule for TPR calc)
>>1300	1	1700
>>1950	0	2350
>>1975	1	1575
>>1960	0.5	1960
>>	    TPR	1896
>
>
>I think this way to calculate performance is wrong.
>
>You should calculate the rating that you need in order not to lose and not to
>win rating from this tournament.
>This should be the performance.
>
>It is not a problem for a computer program to calculate it and you only need the
>exact equation of the rating that can be based on the normal distribution or on
>some linear approximation.
>
>In Israel I remember that the formula was based on some linear approximation(at
>least this was the idea behind the formula in the past).
>
>If my opponent has the same rating I earn 12 points from winning
>If my opponent is 375 points weaker I earn 0.75 points from winning.
>If my opponent is at least 750 points weaker I earn 0 points from winning.
>
>If my opponent is between 0 and 375 points weaker than me the formula is linear
>and the same for difference between 375 and 750 points weaker than me
>
>Examples:
>
>I earn 12 points from winning a player with the same rating
>I earn 11.25 point from winning a player who is 25 point weaker than me
>because 11.25=12-25*3/100
>
>I earn 10.5 points from winning a player who is 50 points weaker than me
>because 10.5=12-50*3/100
>
>I earn 9 point from winning a player who is 100 points weaker than me
>because 9=12-100*3/100
>
>I earn 0.75 points from winning a player who is 375 points weaker than me
>because 0.75=12-375*3/100
>
>between 375 and 750 the formula is also linear but different when 750 elo
>difference say that I earn nothing and 375 difference say that I earn 0.75
>It means for example that I earn:
>0.7 points from winning if the difference is 400
>0.6 points if the difference is 450
>0.5 point if the difference is 500
>
>There was another rule that players get 0.25 point from every game.
>This was the situation many years ago.
>
>There were some changes in the rules and I do not remember the exact rules today
>but I think that they are still based on a linear formula.
>
>Uri

Here is a useful table, for linearly approximating USCF (ELO-based) rating
changes.  The USCF does not use linear approximations, so this table is close
(approx 0-2 points from real rating changes) but not exact.

Also, I believe the USCF rating system guarantees a minimum of 1 rating point
change per win or loss (but I'd have to check to see if this is still true).

Note that the k factor changes to 24 for players of 2100+ strength, and the k
factor changes to 16 for players of 2400+ strength (if my memory is correct), so
this table would look different for players in those stronger rating groups.

To convert to a linear approximation table for players of 2100+ strength,
multiply the shown rating changes by 2/3 (k factors ratio 24/32).

To convert to a linear approximation table for players of 2400+, multiply the
shown rating changes by 1/2 (k factors ratio 16/32).

Here's the basic table:

USCF Rating Changes--Linear Approximation for estimating, for k factor = 32 (max
points won/lost per single game), for players under 2100 rating.
	Higher Rated Player		Lower Rated Player
Rtg	Rating Change  -----            Rating Change  -----
Delta   Win	Draw	Loss		Win	Draw	Loss
0	16	  0	-16		16	0	-16
25	15	 -1	-17		17	1	-15
50	14	 -2	-18		18	2	-14
75	13	 -3	-19		19	3	-13
100	12	 -4	-20		20	4	-12
125	11	 -5	-21		21	5	-11
150	10	 -6	-22		22	6	-10
175	9	 -7	-23		23	7	 -9
200	8	 -8	-24		24	8	 -8
225	7	 -9	-25		25	9	 -7
250	6	-10	-26		26	10	 -6
275	5	-11	-27		27	11	 -5
300	4	-12	-28		28	12	 -4
325	3	-13	-29		29	13	 -3
350	2	-14	-30		30	14	 -2
375	1	-15	-31		31	15	 -1
400	0	-16	-32		32	16	  0




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.