Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Programmers: Best direction to further develop Fruit code?

Author: Uri Blass

Date: 13:42:38 06/20/05

Go up one level in this thread


On June 20, 2005 at 16:09:53, Dann Corbit wrote:

>On June 20, 2005 at 15:42:41, Uri Blass wrote:
>
>>On June 20, 2005 at 15:28:16, Dann Corbit wrote:
>>
>>>On June 19, 2005 at 17:58:07, Uri Blass wrote:
>>>
>>>>On June 19, 2005 at 15:18:33, Vincent Diepeveen wrote:
>>>>
>>>>>On June 19, 2005 at 02:24:52, Roger D Davis wrote:
>>>>>
>>>>>>Since Fruit is just a notch down from Shredder and Fritz, it looks like that
>>>>>
>>>>>Fruit is lightyears away from Shredder.
>>>>>
>>>>>something that might play well in 1 0 because it gets instantly 11 or 12 ply,
>>>>>doesn't mean it is having a chance in some serious time control like 60 in 2
>>>>>under tournament conditions.
>>>>
>>>>I do not know if Fruit has problems at long time control(results at 40/40
>>>>adapted to speed of PIV 2 ghz suggest that it is number 2 after shredder at
>>>>longer time control than 1 minute per game at least if you use nunn type match)
>>>>but if it has problems the logical reason that I can see is bugs in the search
>>>>that prevent it to go deep enough.
>>>>
>>>>I already mentioned one bug of extending single replies too much so fruit can
>>>>solve the famous mate in 30 in 1 seconds and probably cause it to extend often
>>>>useless lines in games).
>>>>
>>>>Fabien wrote that there are bugs in the search that he still had no time to fix
>>>>so he probably meant that there is more than one bug.
>>>>
>>>>It is possible that Fruit is weaker at long time control(not in nunn type games)
>>>>because of positional learning of shredder but I guess that the part of
>>>>positional learning can be done by fabien.
>>>
>>>There is very limited data, but the results I have seen seem to indicate that
>>>fruit is even stronger at slower time controls.
>>
>>1)I see that I made a mistake in my post
>>I meant to say that fruit can solve the mate in 30 at depth 1(not in one
>>second).
>>
>>I think that it is wrong and may cause problems espacailly at long time control
>>because my guess is that these single reply extensions(instead of partial
>>extensions) increase the branching factor of fruit with no significant benefit.
>>
>>2)The data that I see based on reading posts give me no information if fruit is
>>better at long time control.
>
>Did you see the 90 minute plus 30 second contest?

Yes

I saw fruit is leading 4-1 against Shredder9(and fruit is losing the 6th game)
but there are more games at 40/40 time control at CEGT and in this tournament
Fruit is only number 3 at this moment after Fritz.

I also saw results when Fruit beated Fritz at blitz
see http://wbforum.volker-pittlik.name/viewtopic.php?t=2867

Fruit-Fritz 52.5-47.5

I can only say that nothing is clear based on the results.

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.