Author: Dann Corbit
Date: 13:10:46 12/29/04
Go up one level in this thread
On December 29, 2004 at 15:38:25, Uri Blass wrote: >On December 29, 2004 at 15:18:31, Dann Corbit wrote: > >>On December 29, 2004 at 14:58:43, Dann Corbit wrote: >> >>>On December 29, 2004 at 14:53:42, Uri Blass wrote: >>> >>>>On December 29, 2004 at 14:43:40, Dann Corbit wrote: >>>> >>>>>Here is the current result of a little 10'+10" contest I have running: >>>>> Program Elo + - Games Score Av.Op. Draws >>>>>1 Fruit_src : 2850 66 149 52 83.7 % 2566 17.3 % >>>>>2 Crafty-1917 : 2670 85 86 52 61.5 % 2588 26.9 % >>>>>3 Glaurung : 2660 85 89 53 60.4 % 2587 22.6 % >>>>>4 Arasanx : 2648 90 90 52 57.7 % 2594 19.2 % >>>>>5 Movei : 2572 81 93 52 45.2 % 2606 25.0 % >>>>>6 Ges : 2518 82 83 53 37.7 % 2605 30.2 % >>>>>7 Natwarlal : 2494 100 80 52 33.7 % 2612 21.2 % >>>>>8 GreKo : 2389 147 69 52 20.2 % 2627 13.5 % >>>>> >>>>>Conclusion: >>>>>Fruit 2.0 looks pretty strong. >>>>>The Ges version is 1.32, and I was a bit surprised at how well Ges performed. >>>> >>>>I am surprised by the bad result of movei. >>>>It is certainly weaker than Fruit and probably also weaker than Crafty inspite >>>>of the fact that I do not think that it is weaker than Crafty19.17 that is known >>>>to be a weak version of Crafty but I think that it is stronger than Glaurung and >>>>Arasan. >>>> >>>>public version even performed better than Crafty19.15 in a recent tournament >>>> >>>>see >>>>http://www.playwitharena.co.uk/modules.php?name=ATL_League >>> >>>I think that there may have been a problem in recoginzing the movei book. I >>>will look into it. >> >>Movei, Fruit, and Natwarlal are not using a book -- not sure why yet. >>In a 10'+10" contest, lack of a book should be very telling. Not sure why Fruit >>should still be pounding the stuffings out of everybody. > >In the case of movei you need to change the name of the book based on the >version number of the program > >Version number is given in defs.h >I guess that the version number that you have is 300(no significant difference >in playing strength from 295 expected) > >A book is important for not losing the same game again and again. > >Fruit simply does not lose when you choose weak opponents for it so it does not >need book. > > >It is not the first time that a program does good results with no book and >List is leading the rating list of the infinite loop with no book > >http://freehosting.hostrave.com/p/iggor/nil/NIL7_rat.txt > >books are simply not very important(at least books that are used by amateurs) >and it seems that the only importance of them is to prevent losing the same game >again and again. I suspect with a time control of 10'+10" the book will be quite important. I do not have data to prove it yet, but by observations, I see a 40% or so time advantage for the base when a book is used in this format of time control. Suppose (for instance) that program 'A' can find ten moves in the book and program 'B' does not use a book. Then program A will have 100 seconds bonus (no time needed for book moves) and program B will lose the time needed to perform 10 moves (perhaps 4 or 5 minutes as the first move often gets a big time bonus for being the first move). That means that program A will have a much larger "disaster bank" for a fail-low than program B does.
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.