Author: Vincent Diepeveen
Date: 04:56:34 10/11/02
Go up one level in this thread
On October 10, 2002 at 03:05:28, Tony Werten wrote: nullmove improves branching factor, so it isn't 2 ply. The more efficient you normally search, the more it improves branching factor, the bigger the difference when you get a bigger amount of nodes. 4.2 (my branching factor with hashtables and fullwdith) versus branching factor 2.8-3.2 (branching factor what most get with nullmove) is going to pay off each ply extra you get. >On October 09, 2002 at 23:06:48, Robert Hyatt wrote: > >> >> >>I had to stop the experiment sooner than I wanted, but I did find some >>interesting things out. >> >>1. at _very_ fast time controls (40 moves in 1 minute) null-move completely >>destroys non-null-move >>by a ridiculous margin. (this ended something like 60 wins, 5 losses, 8 draws) >> >>2. At longer time controls (40 moves in 10 minutes) non-null-move catches up >>somewhat. It still loses >>far more than it wins, but not _nearly_ so bad as test 1. (this was closer, but >>with fewer games played) >> >>3. At 40 moves in 60 minutes, things close up even closer as the data I >>previously posted shows. The >>margin ended up at 13.0 / 22.0 in favor of null-move. (null won 8, lost 4, >>drew 10) >> >>4. I wonder what would happen at longer time controls. Note that these were >>run on 550mhz processors, going to cpus 4x faster might close the gap even >>farther... >> >>food for thought... >> >>I don't know what 13/22 turns into, rating wise, 18/24 would be +200, so this >>is significantly less than >>+200 overall. > >Can't say I'm suprised. If nullmove gives 2 ply (wich is what is generally >accepted ) then you're basicly doing (n vs n+2) matches wich get less unequal >with bigger n. > >Tony
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.