Author: Christophe Theron
Date: 12:31:11 03/09/04
Go up one level in this thread
On March 08, 2004 at 15:03:10, Chessfun wrote:
>On March 08, 2004 at 05:27:38, Christophe Theron wrote:
>
>>On March 07, 2004 at 11:46:26, Chessfun wrote:
>>
>>>
>>>Minimum 70 games to qualify.
>>>All 3,4 and 5 piece tablebases.
>>>Native books where made, general.ctg where not.
>>>Majority 9100+ games over 60 mins.
>>>Games are autoplayed ponder=on and single PC ponder=off.
>>>AMD XP 2600+
>>>
>>>Hiarcs8 Bareev games at 60' ponder=off using Hiarcs9.ctg
>>>
>>>Currently testing Ruffian 2.1 with Lokasoft book.
>>>
>>>
>>> Program Elo + - Games Score Av.Op. Draws
>>>
>>> 1 Shredder 8 : 2721 44 47 150 67.3 % 2595 37.3 %
>>> 2 Deep Fritz 8 : 2684 45 42 170 61.2 % 2605 36.5 %
>>> 3 Hiarcs 9 : 2640 46 37 190 55.5 % 2602 35.3 %
>>> 4 Fritz 8 : 2633 29 23 470 55.9 % 2592 38.5 %
>>> 5 Deep Fritz 7 : 2633 32 26 374 56.3 % 2589 36.6 %
>>> 6 Deep Junior 8 : 2630 43 39 210 57.1 % 2580 28.6 %
>>> 7 Tiger 15 normal style : 2629 33 24 377 54.9 % 2594 42.4 %
>>> 8 Shredder 7 : 2629 37 31 290 55.3 % 2591 32.8 %
>>> 9 Tiger 15 gambit agres : 2626 48 44 163 58.0 % 2570 31.3 %
>>
>>
>>
>>Very interesting statistics.
>>
>>I have just noticed the striking difference in number of draws between CT15
>>normal and CT15 gambit aggressive.
>>
>>They are close in strength, but the difference in the draw percentage is BIG.
>>
>>I would not say it's surprising, but it's a confirmation for me.
>>
>>The smallest number of draws in this list is achieved by Deep Junior 8.
>
>The list is the top 20 only Junior 7 has the same number.
>23 Junior 7 : 2576 30 23 528 50.1 % 2575 28.6 %
>
>Although you would also expect Junior 8's number to be very close but in this
>case that hasn't happened. I remember during beta testing one of the Tiger's Jim
>Walker and I commented on this fact regarding Junior 7.
>
>I think Gambit Agressive plays to win and sometimes this results in a loss
>sometimes a win, maybe against humans this trying to win would result in a
>higher rating than the default.
>
>Also if you look at both Tiger 15 and 14 the draw percentage is very close.
Over the years I have received numerous comments about the high drawing rate of
Chess Tiger (default style) and I thought the only way to improve on this
(because I agreed it would be better to have a smaller draw percentage) was to
keep on doing what I did all the time: make the program play better, so some
draws would simply be converted to wins.
I looks like playing better also helps to convert some losses into draws, so the
end result is that the draw percentage is almost unchanged! :)
I notice that one of the ways to have less draws, and it was not so obvious to
me, is to boost the "aggressivity" of the program. Statistically, more of the
games that were drawn will either fall on the loss or on the win side.
"Aggressivity" in the case of CT15 aggressive is the weight given to King
attacks.
Now if I could only find a way to detect those attacks that are going to fall on
the "loss" side! :)
Christophe
>>It looks like the strong speculative programs are the ones with a low draw
>>percentage. It was expected, somehow, but it is always very interesting to have
>>a confirmation by numbers.
>
>I would say I'd agree with that, probably similar even with Human chess.
>
>Sarah.
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.