Author: Michel Langeveld
Date: 21:11:25 03/14/04
Go up one level in this thread
On March 14, 2004 at 15:40:43, Bert van den Bosch wrote:
>On March 14, 2004 at 12:31:49, Michel Langeveld wrote:
>
>>On March 14, 2004 at 06:58:50, Bert van den Bosch wrote:
>>
>>>I was wondering when you can decide to make a move which was computed at a ply n
>>>alpha beta, knowing that a 'move now' came while searching an n-ply alpha beta.
>>>You can also make the move computed at ply n-1 (the previous alpha beta tree in
>>>the iterative deepening proces), but I guess sometimes you have found a pretty
>>>nice move at the ply n tree. Is it a matter of taste?
>>>
>>>greetings Bert
>>
>>If the 2nd move is stronger as the first move when search N ply I update the
>>move to play. If time is finished and ply N is not totally calculated to the end
>>it knows that the 2nd move is the strongest one.
>
>
>But this second move can be a bad move, maybe with some bad luck losing a queen
>or so, while the move at ply n-1 doesn't.
It works as follows:
Ply 1
Move 1/4 d4 = 0.40
Move 2/4 e4 = 0.56 [move is sorted higher-new movelist is e4 d4 f4 Qxf7]
Move 3/4 f4 = 0.56
Move 4/4 Qxf7+ = 2.56 [move is sorted higher-new movelist is Qxf7 e4 d4 f4]
Ply 2:
Move 1/4 Qxf7 = -4.00 (Qxf7 is searched first, because this was the best move)
(if the time is up now ... the time is extended]
Move 2/4 e4 = 0.54
[time is up - and not all move have been examined]
Now e4 will be played
The extending if time can be quite ggressive. Some programs can use 1/4 of their
time. I think it's even legal to play directly move e4, because this move was a
lot better as -4.00 at ply 1. Searching futher is also possible, because Ply 2
seems to be an important play. If the time is up before e4 can be found on ply 2
then a program has to play e4! This is a timemanagement issue I think in each
program it is implemented different.
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.