Author: Stan Arts
Date: 06:55:00 01/14/04
Go up one level in this thread
Hoi Michel I wonder, does TSCP, or your version have an aspiration search at the root? If not, do you get similair large differences between PVS and regular alpha- beta when allready using an aspiration window? What I mean is, if you do not generate captures in q-search that won't raise the score anywhere near enough, then I imagine there'll be a lot more nodes in q-search with regular alpha-beta with infinite window, then with the minimal windows PVS creates in the searchtree ? So my question is basicly if most of the savings from using PVS in your case come from less nodes spent in q-search or mostly elsewhere. Greetings, Stan On January 13, 2004 at 10:15:20, Michel Langeveld wrote: >Just wanted to share a number. >I tested TSCP with and without PVS. > >The difference is : 24.3% - 39.0% > >To be precise the TSCP version I used has: >- killers >- history killers >- alfa beta >- no hashtables >- no nullmove > >Opening position Gothic Chess: >rnbqckabnr/pppppppppp/10/10/10/10/PPPPPPPPPP/RNBQCKABNR w KQkq - > >without PVS > >6 215 305037 -10 b1c3 b8c6 i1h3 i8h6 e2e4 e7e5 >7 1575 2511295 18 b1c3 b8c6 i1h3 i8h6 e2e4 e7e5 f2f4 > >with PVS > >6 190 261471 -10 b1c3 b8c6 i1h3 i8h6 e2e4 e7e5 >7 1192 1889545 18 b1c3 b8c6 i1h3 i8h6 e2e4 e7e5 f2f4 > >boostup of 24.3% > >Tactical position in Gothic Chess (38.4%) >r1q2r2k1/p1p1p3pp/10/5p1p2/2P2P3P/2PC3PK1/P2B1Nc3/1R1Q5R w - - > >without PVS >6 595 903139 -486 i3j3 h5h4 d1j7 i8j7 j3i4 g2h2 i4h4 f8d8 j1d1 d8d3 >d2d3 >7 4065 6253437 -9990 i3j3 h5h4 d1j7 i8j7 j3i4 c8e8 i4i5 g2i2 i5j5 e8h5 > >with PVS >6 525 817502 -486 i3j3 h5h4 d1j7 i8j7 j3i4 g2h2 i4h4 f8d8 j1d1 d8d3 >f2d3 >7 2481 3947446 -9990 i3j3 h5h4 d1j7 i8j7 j3i4 c8e8 i4i5 g2i2 i5j5 e8h5 > >boostup of 39.0% > >For the source read next post
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.