Author: Robert Hyatt
Date: 08:07:35 07/17/04
Go up one level in this thread
On July 17, 2004 at 07:59:56, Jan K. wrote: >hi, until this moment i've been using very general search based on PVS+hash+null >move+some other basic things. Now i wanted to improve it but i need help with >this kind of behavior: >from the starting position after e2e3: >5 0.12 1099 0.008 b8c6 b1c3 g8f6 g1f3 d7d5 >6 -0.15 2776 0.013 b8c6 b1c3 g8f6 g1f3 d7d5 d2d4 >7 0.00 10701 0.030 b8c6 b1c3 g8f6 g1f3 d7d5 f1b5 c8d7 >8 -0.27 111778 0.301 b8c6 b1c3 g8f6 g1f3 e7e5 d2d4 e5e4 f3g5 >9 -0.03 577204 1.366 g8f6 g1f3 e7e6 b1c3 b8c6 >10 -0.26 1497107 3.286 g8f6 g1f3 e7e6 f1e2 b8c6 e1g1 f8e7 b1c3 e8g8 d2d4 >11 -0.02 3684836 8.146 g8f6 g1f3 e7e6 f1e2 f8b4 e1g1 e8g8 a2a3 b4e7 b1c3 b8c6 >12 -0.24 30007224 47.015 e7e5 g1f3 e5e4 f3d4 g8f6 f1e2 f8e7 b1c3 e8g8 e1g1 b8c6 >d4f5 > >Note the time difference between depth 11-12. At the root node of "after-e2e3 >position" depth=12 the pv move from previous iteration gives a value and when >searching with (-alpha-1,-alpha) only e7e5 falls between (alfa,beta) and >therefore there's one research to depth 11-obviously it takes 4x more time >compared to the previous search to depth 11. This may be due to bug i'm unable >to find, but still-what is the way to handle these situations when you get >stable pv from depths 1-x and then suddent change at x+1? For me it takes lots >of time even in rather simple positions as the one stated above.... > >thx for any advice :) Nothing you can do. The PV move generally takes over 1/2 the time for an iteration. If you change your mind and choose a new best move, you double that since you do it twice. Your effective branching factor goes down the tubes. If you change your mind 3-4-5-6 times, you can expect it to be worse. the 2-3 effective branching factors occur when there is no PV move change from one iteration to the next.
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.