Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Number of games to test

Author: Christophe Theron

Date: 08:03:38 03/09/02

Go up one level in this thread


On March 09, 2002 at 02:46:42, TEERAPONG TOVIRAT wrote:

>
>
>I see the correlation between  score ratio and the number of games
>at specific %  of confident interval.
>Would u please calculate this correlation for me?
>
>score          the number of games at 90% confident interval
>60%              ?
>70%              ?
>80%              ?
>90%              ?
>
>Thanks in advance,
>Teerapong



I personally use the following tables. Study them and you will quickly
understand that the number of games needed to draw a reasonable conclusion
exceeds what common sense believes. Common sense sucks on this matter, don't
trust your feelings.

Explanations:

1) I know that assuming 1/3 chances for wins, draws and losses is not correct,
but I think it's close enough to reality and does not invalidate the reliability
of these tables.

2) How to read the tables: for example, if you want 90% reliability in your
conclusions and have played 10 games, then you must assume a +/-20% error margin
in the winning percentage of the winner (which translate to a +/-140 elo margin
of error). So if program A beats program B by 65% in a 10 games match, then you
cannot even tell which program is better. Play more games.

3) These tables should be taken with a statistical grain of salt. So if you
don't understand the concept of margin of error, reliability percentage of a
result and so on, just forget about them and go back to tic-tac-toe. ;)



Reliability of chess matches
(assuming each opponent has 1/3 chances to win, 1/3 to loose and 1/3 to draw)

90% confidence
Games	%err+/-	elo+/-
    10	 20	140pts
    20	 15	105pts
    25	 14	 98pts
    30	 12	 63pts
    40	 10	 70pts
    50	  9	 56pts
   100	  6.5	 35pts
   200	  4.72	 33pts
   400	  3.34   23pts
   600	  2.66	 19pts
   800	  2.39	 17pts
  1000	  2.12	 15pts
  1200	  2.00	 14pts
  1400	  1.81	 13pts
  1600	  1.66	 12pts

80% confidence
Games	%err+/-	elo+/-
    10	 15	105pts
    20	 11	 77pts
    25	 10	 70pts
    30	  9	 63pts
    40	  8	 56pts
    50	  7	 49pts
   100	  5.0	 35pts
   200	  3.75	 26pts
   400	  2.60	 18pts
   600	  2.15	 15pts
   800	  1.86	 13pts
  1000	  1.66	 12pts
  1200	  1.46	 10pts
  1400	  1.40	 10pts
  1600	  1.34	  9pts

70% confidence
Games	%err+/-	elo+/-
    10	 15	105pts
    20	 10	 70pts
    25	  8	 56pts
    30	  8	 56pts
    40	  6.3	 44pts
    50	  6.0	 42pts
   100	  4.0	 28pts
   200	  3.0	 21pts
   400	  2.2	 15pts
   600	  1.7	 12pts
   800	  1.5	 11pts
  1000	  1.3	  9pts
  1200	  1.24	  9pts
  1400	  1.14	  8pts
  1600	  1.04	  7pts



    Christophe



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.