Author: Uri Blass
Date: 02:33:32 03/26/04
Go up one level in this thread
On March 25, 2004 at 23:22:18, Dann Corbit wrote: >On March 25, 2004 at 23:06:01, Mike Byrne wrote: > >>On March 25, 2004 at 21:48:27, ludicrous wrote: >> >>>Are you willing to bet that Chess Genius for Nokia 6600 cellphone will not win a >>>single game, nor even draw one, against Shredder 8 / Deep Fritz 8 with in blitz >>>time settings, in 100,000 games? >>> >>>Just a thought. >> >>It reads like a loaded question - show us the game if you have one. > >It would be a bad bet to take. Why will someone make it? > >Win expectancy table: > >0 points is 0.5 >100 points is 0.359935 >200 points is 0.240253 >300 points is 0.15098 >400 points is 0.0909091 >500 points is 0.0532402 >600 points is 0.0306534 >700 points is 0.0174721 >800 points is 0.00990099 >900 points is 0.00559197 >1000 points is 0.00315231 >1100 points is 0.00177512 >1200 points is 0.000999001 >1300 points is 0.000562025 >1400 points is 0.000316128 >1500 points is 0.000177796 >1600 points is 9.999e-005 >1700 points is 5.6231e-005 >1800 points is 3.16218e-005 >1900 points is 1.77825e-005 >2000 points is 9.9999e-006 >2100 points is 5.62338e-006 <<<<<<<<<<<<<<<<<< Here we are. > >So (for instance) if one of those top dog buggers is 2750 (relatively speaking) >then the phone thingy would only have to be 650 Elo to get even money. > >Now, things do break down in the tails. But the math does not support a total >blanking. This math is clearly wrong. You cannot know the expected result based on the difference in rating. There are cases when 500 elo difference is enough for 100000-0 score and there are cases when 2000 elo difference is not enough. You may even have cases when 0 elo difference is enough for 100000-0 because the players are deterministic and always repeat the 2-0 result for A but when they play against other players you find that they are equal. Uri
This page took 0.22 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.