Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Gauntlet Pharaon 3.1 beta KHS - clear improvement !

Author: Uri Blass

Date: 09:51:26 01/03/05

Go up one level in this thread


On January 03, 2005 at 11:49:59, Jason Kent wrote:

>On January 03, 2005 at 11:29:48, Uri Blass wrote:
>
>>On January 03, 2005 at 11:14:15, Jason Kent wrote:
>>
>>>If Pharaon can score this well at 5' + 5", then it will prove to be a monster at
>>>long time controls.
>>
>>I see no connection.
>>
>>The only way to know if pharaon3.1 can do better at long time control is by
>>testing it.
>>
>>Maybe previous versions can do better at long time control but it does not prove
>>if the last version can do better or cannot do better at long time control.
>>
>>Uri
>
>Uri,
>
>While it is possible the newest "bugfixes" could hurt it's strength, I doubt it.
> Pharaon has had a reputation at excellent results in long time controls for
>quite some time, and I don't think that will change with a couple bug fixes.
>This does look promising to me inspite of your pessimism.  Maybe you are
>jealous?
>
>Jason

I do not claim nothing about long time control with no evidence.
I also think that generally strong programs in blitz are better at longer time
control.

I remember claims that quark is better at long time control but I see no clear
evidence for it based on data.

AEGT rating at 40/4

1 Pro Deo 1.0                    : 2689   19  22   845    62.4 %   2601   25.9 %
2 Ruffian 1.0.5                  : 2679   18  19  1046    60.9 %   2602   26.7 %
3 Aristarch 4.50                 : 2639   19  16  1197    53.6 %   2614   28.6 %
4 SOS 4                          : 2625   21  18   995    52.9 %   2605   26.1 %
6 Delfi 4.5                      : 2617   22  17   990    51.2 %   2608   29.5 %
8 Tao 5.7 b07                    : 2592   18  21   966    47.4 %   2610   26.5 %
9 Crafty 19.15                   : 2587   24  27   530    42.8 %   2637   30.6
%13 Yace 0.99.87                   : 2579   19  22   824    45.6 %   2609   29.4
%
14 Gothmog 1.0 beta 10            : 2564   22  23   741    43.5 %   2609   25.1
%
15 Quark 2.35                     : 2522   24  20   773    37.3 %   2612   25.7
%

AEGT rating at 40/40


1 Ruffian 1.0.5                  : 2679   31  33   334    63.2 %   2585   31.1 %
2 Pro Deo 1.0                    : 2678   39  46   204    64.7 %   2573   26.5 %
3 Aristarch 4.50                 : 2657   32  32   333    60.1 %   2587   30.6 %
5 Delfi 4.5                      : 2638   34  30   324    57.1 %   2588   33.3 %
6 SOS 4                          : 2634   38  47   208    65.1 %   2525   25.5 %
9 Tao 5.7 b04                    : 2603   37  28   336    51.8 %   2590   31.5 %
10 Crafty 19.15                   : 2592   37  26   336    52.2 %   2577   37.2
%
12 Yace 0.99.87                   : 2574   28  37   325    47.2 %   2593   33.5
%
16 Quark 2.35                     : 2551   31  34   335    43.7 %   2595   27.8
%
18 Gothmog 1.0 B10                : 2546   29  37   327    47.4 %   2564   30.0
%


You can see that the order is almost the same and the difference for Quark
between 4/40 and 40/40 is estimated to be less than 26 elo for long time
controland the possible error at long time control is bigger than 26 elo(I
calculated it by head so I hope no mistakes).

Yace 57-23=34
Crafty 65-41=24
Tao 70-52=18
Delfi 95-87=8
Sos 103-83=20
Aristarch 117-106=11
Ruffian 157-128=29
Pro deo 167-127=40
Gothmog 42+5=47
(34+24+18+8+20+11+29+40+47)/9=231/9<26 elo


You can see the rating lists in and I simply took programs that appear in both
lists

http://www.husvankempen.de/nunn/40403012.txt
http://www.husvankempen.de/nunn/4043012.txt

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.