Author: Robert Hyatt
Date: 08:17:42 03/01/06
Go up one level in this thread
On March 01, 2006 at 05:35:42, Vasik Rajlich wrote:
>On February 28, 2006 at 19:50:09, Tord Romstad wrote:
>
>>On February 28, 2006 at 18:05:32, Miguel A. Ballicora wrote:
>>
>>>This one:
>>>
>>>>>> if (nodecount == 15 && depth == 2) {
>>>>>> depth--;
>>>>>> }
>>>Is exactly what Tord mentions (a very simple raw version) applied close to the
>>>tips.
>>
>>In a certain sense, it is exactly the opposite of what I do. I only
>>use reductions *far* from the leaves; more specifically when deph > 2.
>>Before I started using null move threat detection to improve the tactical
>>accuracy, I had to use depth > 3 in order to avoid too many horizon
>>effect problems.
>>
>>Tord
>
>That's one difference. The really big difference was the "if (nodecount == 15)
>break;". That's Shannon B-type search!
>
>Vas
Since those were all surrounded by conditional compilation tests, I assumed that
was an "evolution". First tried just tossing them out (which Bruce and I tried
first too, btw) then later he tried reducing them instead, which was again
exactly what I played around with... I just never got to the idea of not
reducing on moves that had failed high above some fixed percentage of the times
they were tried...
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.