Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Tao 5.6 vs Aristarch 4.41 (15m+10s) after 13 games (6,0-7,0) PGN

Author: Bryan Hofmann

Date: 12:49:02 03/14/04

Go up one level in this thread


On March 14, 2004 at 15:17:26, Uri Blass wrote:

>On March 14, 2004 at 14:36:56, Bryan Hofmann wrote:
>
>>On March 14, 2004 at 13:07:44, Jorge Pichard wrote:
>>
>>>On March 14, 2004 at 12:56:49, Kurt Utzinger wrote:
>>>
>>>I expect Aristarch 4.41 to score better than that against Yin 5.6, since
>>>ElChinito 3.25 and Aristarch 4.41 are almost equal in strength:
>>>
>>>Arena tournament
>>>
>>>Rank	Engine	Score	Yi	                EL                     S-B
>>>1	ELChinito_325	11.5/18	· · · · · · · · 1=100010=111110=11. . .74.75
>>>2	Yin56 	6.5/18	0=011101=000001=00	· · · · · · · · · . . .74.75
>>>
>>>
>>
>>I don't know what length you matches were, but Tao is really not much of a Blitz
>>player, it's strenght is is longer time controls.
>>
>>
>>Bryan
>
>I do not agree
>I think that Tao is strong at all time controls.
>
>For Tao5.4's results in blitz in the nunn positions
>see http://www.husvankempen.de/nunn/ratings/ratingsnunn2.htm
>
>It is clearly better than most programs in blitz and Tao5.6 performs even better
>see http://www.husvankempen.de/nunn/tabellen/nunntog.htm
>
>Tao is probably better than movei at all time controls and movei is not a weak
>blitz program and seems to be at the middle of the rating list.
>
>Uri

I guess this is one time when I glad you do not agree. But, I have been beta
testing Tao for Bas for about a month. I have run matches on FICS (TaoGizmo) and
at home with different settings. What I have discovered is when the time control
is below 10/0 to say 5/0 or lower Tao does not preform as well against the same
engines that it is able to win against in longer time controls. For example, I
tested Tao against Smarthink, in 60/0 games it was 8-2-0 for Tao, in short
matches, 5/0 it was 5-3-2 Smarthink. I am aware of the links you provided above
and Tao is outstanding in these positions, it just a different story in playing
normal games.

I'm still testing different settings at different time controls to see if I can
change this. I can report that using the follow settings in long games with
pondering does very well as I ran them for a week on FICS and quickly jumped to
the top ten systems in standard ratings. It should be noted that running these
settings in matches on a single CPU will not do well as pondering is a must.

PreExtension=24
CheckExtension=24
MateExtension=24
OneRepExtensions=24
RecapExtensionType=normal
RecapExtensions=12
HalveExtensions=2.5
NullMoveType= 2/3

Bryan




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.