Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Something wrong with my futility pruning?

Author: Severi Salminen

Date: 02:09:47 12/06/00

Go up one level in this thread


On December 05, 2000 at 20:24:24, Bruce Moreland wrote:

>
>On December 05, 2000 at 16:46:17, Severi Salminen wrote:
>
>>Hi!
>>
>>Check out this position:
>>
>>[D]rnbqkbnr/ppp2ppp/8/3pp3/3PP3/8/PPP2PPP/RNBQKBNR w
>>
>>With futility pruning my program does this:
>>
>>
>>     Time Depth   Score       Nodes PV
>>       0    1     -0.14         175 Be3
>>    0.00    2     -0.19        1851 Be3 Be6
>>    0.11    3     -0.11        7755 Nf3 Bb4+ Bd2 dxe4 Bxb4 exf3 dxe5 Qxd1+ Kxd1
>>fxg2 Bxg2
>>    0.33    4     -0.14       22227 Nf3 exd4 Qxd4 Be6
>>    0.98    5      0.00       67477 dxe5 dxe4 Qd7+ Qxd7
>>    3.07    6     -0.12      324618 Nf3 Bg4 Be2 Bxf3 Bxf3 Bb4+
>>   12.41    7      0.00     1515184 Qh5 dxe4 Qxe5+ Qe7 Qxc7 Qxc7
>>   33.06    8     -0.09     4203195 dxe5 dxe4 Qxd8+ Kxd8 Bc4 Nc6 Bf4 Bb4+
>>
>>And without:
>>
>>     Time Depth   Score       Nodes PV
>>      50    1     -0.14         175 Be3
>>    0.05    2     -0.19        1851 Be3 Be6
>>    0.16    3     -0.11        7850 Nf3 Bb4+ Bd2 dxe4 Bxb4 exf3 dxe5 Qxd1+ Kxd1
>>fxg2 Bxg2
>>    0.44    4     -0.14       22555 Nf3 exd4 Qxd4 Be6
>>    1.10    5     -0.09       73064 Nf3 dxe4 Nxe5 Bb4+ c3
>>    2.85    6     -0.12      306521 Nf3 Bg4 Be2 Bxf3 Bxf3 Bb4+
>>    9.61    7     -0.07     1212161 dxe5 dxe4 Qxd8+ Kxd8 Bc4 Bb4+ Nd2
>>   29.33    8     -0.09     3947537 dxe5 dxe4 Qxd8+ Kxd8 Bc4 Nc6 Bf4 Bb4+
>>
>>Why the h?%l is the one without futility pruning faster and has less nodes? I
>>use futility margin of 2 pawns.
>>My futility code is something like: (I prune check now also...)
>>
>>if(depth==1)
>>  if(futility_condition(Moves[k].value)<=alpha)
>>    continue;
>>
>>the futility_condition() returns the current material balance + the material
>>gain of the move+ 200.
>>
>>Any ideas?
>>
>>BTW. Does my node counts look reasonable? I'm using MVV/LVA ordering in qsearch
>>+ nullmoves (R=2).
>>
>>Severi
>
>It switched moves in a later ply in the one where you prune, and this costs
>nodes.
>
>It is not possible to determine whether a pruning technique is good or bad based
>upon one position.
>
>Who may as well decide a coin is lucky or unlucky depending upon whether it
>comes up heads the first time.

Yes, of course. That's why I asked. In heinz's book he says that tree sizes are
shrinked about 60% on average, but in my case this did not happen. It would be
nice to have a set of positions with which one could verify these things. Do you
now any futility -positions (in which the futility pruning really matters) and
affects?

Severi



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.