Author: Bruce Moreland
Date: 09:53:39 11/30/00
Go up one level in this thread
On November 30, 2000 at 12:09:02, Leen Ammeraal wrote: >I use a trangular array pv[i][j] to >keep track of the principle variation, >as usual. Each row has length pv_length[i]. >However, it is not clear >to me what to do with this array >when I retrieve a move from my hash table. >Can anyone give me a hint, please? >Leen Ammeraal I'm not quite sure why you are confused, but I'll take this at face value and simply answer it. Your hash table in many cases will just tell you what to search first, and if this happens it doesn't affect the PV at all. If your hash table tells you to cut off, it can tell you to cut off high, low, or with an exact score. If it tells you to cut off high or low, your PV isn't affected. If it tells you to cut off with an exact score, the score attached to the move must have been between alpha and beta. In this case, you can create a new PV, with the move that will cause the cutoff, and a length of 1 (one move in the PV). This would worry me a little bit because hash information is not guaranteed to be good, and in the off-chance that a duplicate key error is affecting your PV, you might have an illegal move in your PV. So what I would do is cut the PV here, simply say that it has zero length as of this point. Another experiment to try is to not allow hash cutoffs if the score is between alpha and beta. This should result in longer PV's, which may result in better ordering on re-searches. bruce
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.