Author: Uri Blass
Date: 00:08:17 04/06/04
Go up one level in this thread
On April 05, 2004 at 10:51:06, Tim Foden wrote: >On April 05, 2004 at 08:26:08, Andrew Wagner wrote: > >> >>I'm taking this intro to calc-based statistics class this semester, and it got >>me thinking. We tend to generate a lot of statistics during a game. In fact, we >>could generate a LOT more. But we do very simple things with these statistics. >>For example, we might add some bell or whistle to the search, notice that it >>doesn't show any obvious increase in NPS or move ordering, or whatever we were >>hoping for, and ditch it. That's an extremely simple analysis of the statistics. >>I'm wondering if anybody has done any more advanced statistical work in figuring >>out what factors are important to winning a game. Here are two examples I >>thought of: >> >> 1.) Using advanced statistics to calculate better values to be used in >>static eval. What if we took a ton of super-gm games, and generated a slew of >>statistics about each game like number of moves white has an isolated pawn on, >>or number of moves black keeps his king-knight, and so on. Then do a >>multi-variate analysis to determine whether any of those things have an effect >>on the probability that white or black will win. Unfortunately, this math is a >>little above me at this point, and maybe someone else has tried it, I don't >>know. It's an interesting idea to think about, though. > >Dave Gomboc has a thesis paper (Ordinal Correlation for Evaluation Function >Development) on generating evaluation values from annotated games. This may be >of interest. > >I've used multi-variate linear regression to generate a set of evaluation >values, but I have some bugs in the implementation (as pointed out by Dave >Gomboc when I spoke to him in Graz), and I haven't tested the values at all. I think that the best formula is not linear. A good formula should use all the information that you calculate and get probabilities for win,draw,loss. probability are very important in computer chess and in every game for extensions and reductions decision(you want to extend when you are uncertain about the result and to reduce depth when you are certain about the result when extreme example is that when tablebases tell you that it is a draw you do not search forward and the remaining depth is not important. If you get +0.02 pawns for white then it is important to know if it is 51% win for white and 49% loss or 98% draw and 2% win for white. Uri
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.