Author: Christophe Theron
Date: 16:42:04 01/24/01
Go up one level in this thread
On January 24, 2001 at 18:51:00, Hermano Ecuadoriano wrote:
>On January 24, 2001 at 16:26:23, Robert Hyatt wrote:
>
>>On January 24, 2001 at 15:37:40, Hermano Ecuadoriano wrote:
>>
>>>On January 24, 2001 at 14:25:39, Richard Heldmann wrote:
>>>
>>>>ERT.04
>>>>Pentium 200MMX 64mb RAM Windows 98
>>>>Crafty 18.1 8mb hash
>>>>Chess Tiger 8mb hash
>>>>
>>>> 12345678901234567890
>>>>1 Chess Tiger 13.0 1½1011½111½11½01½½½1 14.5/20
>>>>2 Crafty v18.1 0½0100½000½00½10½½½0 5.5/20
>>>>
>>>>PGN available on request.
>>>
>>>Thanks.
>>>My testing shows Tiger13 better than Gambit Tiger against Crafty.
>>>My testing also shows that their scores against Crafty improve with
>>>slower time controls, and your data here support that. I have
>>>Tiger13 +13 -6 =14 against Crafty 18.1 on a PII-300 at 5 3, and other
>>>results.
>>>
>>>I think that with larger sample sizes, it could be proven that the Tigers
>>>are relatively better against Crafty with slower time controls, or equivalently,
>>>on faster hardware.
>>>Christophe Theron "himself" doesn't believe this has ever been proven.
>>>
>>>Here's one quick experiment: On your 200Mhz machine, at a time control of
>>>5 3 or faster, Gambit Tiger might lose to Crafty. It scored +12 -12 =7 on
>>>mine.
>>
>>
>>I can tell you absolutely that your data is flawed. Crafty is _absolutely_
>>weaker at blitz than it is at longer time controls. And this is based on
>>_thousands_ of test games, not on a few dozen. The things I do in my search
>>rely on fast hardware, or else long time controls. The faster the game, the
>>worse Crafty will do as its simple q-search will get it into trouble faster
>>since null-move failures happen more often at reduced depth...
>
>O.K. I'll look again and report later.
You need a larger sample in order to draw a reliable conclusion.
I have seen a 0.5-9.5 result recently, which is not statistically relevant.
I would suggest that you play at least 100 blitz games, and then 100 g/1h games,
and we will see.
I have no reason to believe that Tiger, or any other program, will perform
better at longer time controls against Crafty, unless the branching factor of
the program in question is better than Crafty's. And I don't think that Crafty's
branching factor sucks. So...
Christophe
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.