Author: Christophe Theron
Date: 10:52:55 03/17/04
Go up one level in this thread
On March 17, 2004 at 12:35:41, Uri Blass wrote:
>On March 17, 2004 at 11:43:12, Anson T J wrote:
>
>>On March 17, 2004 at 10:40:30, Bigler David wrote:
>>
>>>Hi,
>>>
>>>Shredder 8 performed 82 ELO point better than Junior 8
>>>Junior 8 performed 82 ELO point better than Shredder 7.04
>>>
>>>So Shredder 8 performed 164 ELO better than Shredder 7.04
>>>
>>>rgds
>>
>>I see, I thought when it was -41 / +41 that A was 41 weaker than b and b was 41
>>stronger than a. Not that a was 82 weaker and b was 82 stronger. Perhaps you are
>>correct. thx
>
>56% is less than 50 elo difference based on fast calculation.
>
>If the result is 56-44 then I get the following difference in rating
>
>(56-44)/100*400=48 so my calculation give 48 elo.
>
>
>The formula that I use is only approximation to the real formula so I guess that
>41 is correct and not 82.
>
>Uri
There is a simpler formula that is easy to remember.
1) take the winning percentage of the winner
2) if it is above 80% stop here (the formula below does not apply)
3) subtract 50 from the winning percentage
4) multiply by 7
5) the result is the elo difference between the two
In this particular case:
1) winning percentage of the winner is 56%
2) it's below 80%, so I can use my formula
3) 56 - 50 = 6
4) 6 x 7 = 42
5) the approximate strength difference is 42 elo points
DOES THE MATCH DEMONSTRATE THAT ONE OPPONENT IS SIGNIFICANTLY BETTER?
I can add from my tables that the error margin for a 50 games match is, with 90%
confidence, 9%.
So it is not possible to say with a 90% confidence that Shredder 8 has
demonstrated that it was stronger than Junior 8 in this match.
Actually it is not even possible to say with 70% confidence that this match has
demonstrated the superiority of Shredder 8. Error margin for a 50 games match
with 70 confidence is 6%.
So from this result alone, if you want to claim that Shredder 8 is superior to
Junior 8 under the conditions of this match, you have more than 30% chances to
be wrong.
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.