Author: Jouni Uski
Date: 21:57:25 04/22/02
Go up one level in this thread
On April 22, 2002 at 15:03:29, Dann Corbit wrote: >On April 22, 2002 at 09:01:30, Jouni Uski wrote: > >>1. program X plays relatively better against humans than against computers > >Not enough evidence to say, either way. It is easy to get hundreds of games >with computer/computer matches, but computer verses human it is difficult to >collect a large volume of quality data under controlled conditions. > >>2. program Y plays relatively better with longer time control > >This is surely true for some programs. Amy at G/1 minute will be clobbered by >TSCP. Amy at G/5 minutes will be blasted by Wildcat. At G/120, Amy will >destroy them both. On the other hand, there are definitely programs where the >speed of play makes little or no difference in the quality of play. > >At some point, the branching factor of any chess program will dominate the >quality of its play. From that point forward, additional time will make no >difference in the quality of play against opponents where their branching factor >has also began to dominate. > >>3. there are diminishing returns from speed doubling, when search depth is >> increased > >There are conflicting results for this. I don't think we have enough data to >say. > >>I think definitely 3. is true - only the size of diminishing is unclear. E.g. >>from latest SSDF list (yes comp-comp play!) doubling gives only 40 points at >>current top level against 75 previously. But 1 and 2 are still open cases. > >The error bar is far larger than the measurement, in this case. ??? Are You serious? The value of 40 is based on 1000s of games! Error bar is non-existent. Jouni
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.