Author: Andrew Dados
Date: 13:13:04 02/22/06
Go up one level in this thread
On February 22, 2006 at 15:47:31, Dann Corbit wrote: >On February 22, 2006 at 15:45:12, Dann Corbit wrote: > >>It would be nice to know if a particular opening position underperforms or >>overperforms. >> >>By that I mean, given the Elo of the players that reached the position, and the >>Elo of the opponents that reached the position, did the position for the side to >>move exceed expectations or fall below expectations? >> >>I don't think it is as simple as taking the average Elo of the side to move and >>the average Elo of the opponent side, and then examining the win percentage and >>seeing if it meets the Elo estimate within error bounds. >> >>I guess that we would have to examine all of the pairs. >> >>Does any statisical genius have a solid model for this analysis? > >Suppose that we have some opening position and it scores 40%. Sounds pretty >bad. > >But now suppose that the position has 100 games and is always played by an >engine 400 Elo below the opponent. In this case, instead of underperforming it >is overperforming stupendously. Just looking at won/loss/draw is not nearly >good enough. Now suppose some highly tactical opening gets 50% for ratings around 2200 but 35% for 2600+. Just to note one needs to be careful with statistics here. But anyhow the idea is interesting. On a related note - years ago I wrote simple perl script to calculate percentage of draws depending on avg opponent rating as well as on rating difference and results were interesting. So bad I lost that script. -Andrew-
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.