Author: Ricardo Gibert
Date: 15:35:26 09/14/99
Go up one level in this thread
On September 14, 1999 at 16:05:47, KarinsDad wrote: >On September 14, 1999 at 15:38:10, Robert Hyatt wrote: > >[snip] >> >>If you are 400 rating points worse than crafty, you expect to score about 1 >>point out of every 16. Which could be 2 draws or 1 win. A 1700 player, if I >>had to guess, is at least 600 points worse, which would mean 1 win of every 64 >>games. Or two draws. Note that this is a rough estimate of 200 points meaning >>a 1-3 result, which is close but not exact. > >I think that the Elo equations do not quite correspond accurately to >human/computer games. A major reason that a human player 600 points below >another human player can draw or win once in a blue moon is that the higher >rated player is "having a bad day" (stress, illness, fatigue, overconfidence, >distractions, whatever). > >However, programs do not have a bad day in that sense. > >I would think that a program rated 600 points above a human would almost always >(if not always) win against the human. Now, I am sure that some people will >disagree with this viewpoint, however, I think it is fairly accurate. I would be >surprised if a 1700 player could draw 1 game in 1000 from your Quad Xeon Crafty >(assuming full strength Crafty with EGTBs, no takebacks, etc.), even in standard >time controls. > >I think the Elo equations are a closer approximate the closer a human and >computer are in playing strength since I think that programs are a LOT more >consistent in playing strength than most humans are (especially 1700 strength >humans). But, I do not think we have enough data yet (i.e. wide range of >human/computer games) in order to modify the Elo equations for humans vs. >computers. > >KarinsDad :) The Elo equations are based on a number of assumptions and are not intended to be a realistic model. This is particularly so when you consider the "extrema". There it is ridiculous to apply the elo equations, but that doesn't seem to stop people. A pawn is worth about 200 rating points, so the stronger player would have to have a pretty bad day like blow a piece for nothing. To give you an idea, I used to mess around in skittles by spotting A-players the following handicap: They play white and I start the game with my K on e4 (or d4 I can't remember which, it makes a big difference). Believe it or not I'm a heavy favorite! A very instructive hustle.
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.