Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: F6a v F6b 5 mins

Author: Uri Blass

Date: 23:55:05 09/03/00

Go up one level in this thread


On September 03, 2000 at 22:00:59, stuart taylor wrote:

>On September 03, 2000 at 21:48:44, stuart taylor wrote:
>
>>On September 03, 2000 at 21:14:43, Chessfun wrote:
>>
>>>
>>>For what it's worth:
>>>
>>>Cel 433
>>>Ponder=off
>>>Hash 16 Mbs
>>>F6 general.ctg
>>>Tablebases On.
>>>
>>>Fritz6a W16 L24 D34 = 33/74
>>>Fritz6b W24 L16 D34 = 41/74
>>>
>>>Thanks.
>>
>>That's a good sign! It's clearly atleast a little bit better at blitz (unless it
>>is because of better opening book). But hopefully, the difference is even bigger
>>in longer games, i.e. better knowledge.
>>Will skil/knowledge be increased even more for Fritz 7?
>>S.Taylor
>
>I wish I could edit posts which were already sent!
>I'm trying to see if Fritz is clearly, or still will be the over 100 elo better,
>as was claimed. And being superior over its own previous version, isn't enough
>evidence, yet. After all, it still lost a good few, even though it could have
>been expected to know itself well enough, not to.
>S.Taylor

I did not say that Fritz is over 100 elo better but that the result (+9,=7,-3)
suggests that it is over 100 elo better

The last result suggests that it is less than 100 elo better.

400*(24-16)/74=3200/74=43+(9/37)=43.24.. points difference.

This result also can happen if you play the same engine against itself.

If you combine these 2 results you get 33-19 that is significant result based on
who is better but it is not fair to ignore another result that was posted that
was 3:2 without draws for fritz6a on 3 minutes/game so you get 35-22 for fritz6b
that is again not significant by who is better.

You can say that it is not fair to include 5 minutes/game and 3 minutes/game on
the same package but there was no decision to include only 5 minutes for game
before the experiment began so I must include all the games.

The final result from all the posts is 35:22(42 draws) that suggests
400*13/99=52.525252525252... elo difference

I believe that the difference is smaller(I read that there is only a small
change between Fritz6a and Fritz6b).

Uri



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.