Author: Albert Silver
Date: 13:41:34 11/14/04
Go up one level in this thread
On November 14, 2004 at 09:45:05, Mike Byrne wrote:
>On November 14, 2004 at 08:07:39, ThatsIt wrote:
>
>>Hi to all !
>>
>>Because of Eds remarks about my former experiment
>>(i cannot find the link now)
>>i've done it once more, this time with 40moves/8min. + 40/8 + 40/8...
>>and 80 startpositions (write protect) = 160 games per match.
>>Both, ruffian and list, don't produce learnfiles.
>>64MB HTs, 4erTBs, ponder=off
>>
>>1. Match ruffian 2.1.0 vs list 5.12
>>75.0 - 85.0 (50-50-60)
>>
>>After that the procedure was always the same:
>>i moved the databasefiles onto a floppydisk and delete
>>them on the harddiskdrive.
>>Then i shutdown the machine and wait for exact 5 minutes.
>>Afterwards i turned on the machine, wait until the pc was
>>ready, doubleklick the CB-Symbol (first engine which were
>>loaded on the start were always crafty 19.13).
>>Now the tournament with the two engine has been started,
>>ruffian as fist engine, list as the second.
>>
>>repetition # 1
>>78.5 - 81.5 (46-65-49)
>>no great difference in the final result but look at
>>the changes in the win-draw-lost statistics !
>>
>>...same procedure...
>>
>>repetition # 2
>>83.5 - 76.5 (52-63-45)
>>now the match result has turned !!
>>
>>...same procedure...
>>
>>repetition # 3
>>77.5 - 82.5 (47-61-52)
>>the match result has turned again!!
>>
>>Best G.S.
>
>Interesting -- from the way you contructed the match, one would think the
>results wold be identical. So clearly , the only difference has to be variable
>engine time. Which engine is being favored - first or the second, I am not
>sure by your post.
Well, it shouldn't be too hard to get the odd second's difference in a move, but
in that time control, it can be the difference between a new move found, or an
extra ply that is searched. I suspect that with a slower time control, the
results would vary less, and the games would reproduce themselves more often.
Albert
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.