Author: Vincent Diepeveen
Date: 05:32:39 12/03/99
reposted as will perhaps didn't check his email. In Reply to: Re: ICC Green List - Nov 29 posted by Will Singleton on November 29, 1999 at 23:51:32: On November 29, 1999 at 23:51:32, Will Singleton wrote: >On November 29, 1999 at 17:45:38, Vincent Diepeveen wrote: > >>On November 29, 1999 at 16:48:43, Will Singleton wrote: >> >>>Well, if it's any consolation, my program rarely gets more than 6 ply in the >>>middlegame at blitz time controls. I get killed all the time tactically. >>> >>>Will >> >>right. why make all kind of rating graphs and even invent a logarithmic >>function? >> >>Let's get a random value and make a rating graph out of it. > >Well, that's a good point. I often notice that the outcome of certain matches >seems very random-like. And it is troubling. > >But then I think, well, let's say that you and I improve our search functions to >allow 10 ply in blitz. Will we not have similiar results as the 6 ply versions? Definitely not. Let's bet for money if you want to, but i'll at least challenge you to proof your obvious wrong conclusions! Current DIEP version, so no tricks, no extra extensions turned on as the ones that are currently turned on (all are out except check and a passed pawn extension), limited at 6 ply against amateur at 6 ply. So limiting *all moves* it to 6 ply, including endgame. then DIEP at 10 ply against amateur at 10 ply. No pondering as we play with near to infinite time. If you want to i can also run everything on a single processor, so that transpositiontable luck because of parallellism at 10 ply is not gonna get blamed for the higher % afterwards by some scientific dudes that agree with you. Just searching n ply against n ply, without tricks as suggested some years ago like modifying eval to be an eval that did a 4 ply search. Just a 2 minute change that turns off time-check in program and forces it to use the depth=n setting. What i predict is a much closer to 50% score for the 6 against 6 ply match, and a complete annihilation of amateur when both searching at 10 ply. I assume here already of course that DIEP's eval is better than Amateurs. If i would consider the evaluation of amateur a lot better then i would obviously predict the opposite outcome (amateur annihilating diep at 10 ply and nearer to 50% score at 6 ply). Of course things like learning and such must be turned off. Preferably both versions playing the same openings. However by just playing a large quantity of games we can measure statistically accurate what has happened. To make the experiment even better we should log the evaluations too and see where things went wrong and calculate whether that had statistical influence on the outcome. Also we could start with the same book if you want to. > Experimentation has shown that chess is, or perhaps is, infinitely deep. That I am not aware of experiments pointing in the direction you indicate here. I am aware of the opposite findings though. First many tens of years ago De Groot concluded that the majority of tricks are not deeper than about 6 moves (12 ply). Most positional plans are about 6 moves at maximum. There are other less obvious insight for this, look at a huge testset like win at chess. Majority of tricks here are small depth tricksl. Only very FEW need a bigger depth. >is, if you search 14 ply on average, and I search 13, that you will win a >certain percentage more games. So the question returns to the fundamental one: >can you search faster and deeper than your opponent, while retaining enough >positional knowledge to avoid unplayable positions? Or maybe, can you limit >your wonderful positional eval enough to allow a deep enough search not to get >killed tactically? I think last few years has shown obviously that the opposite is true. Each tactical trick can be seen at a certain depth for a program (hopefully). Don Dailey has done some experiments which indicate the obvious, namely that a match with programs searching at n and n-i , i > 0 depths with different n and i, doesn't show the same difference in % anymore but a lot less when n gets bigger. If i search say 14 ply or something, then i'm quite sure that when playing a lot of different opponents which search 13 ply, the score is not gonna be as well as a match where you search 6 ply against opponents searching 5 ply. There are however a lot of obstacles we must keep in mind a) some programs have a much better and especially bigger evaluation >So keeping track of the progress of these developmental programs merely shows >who has been more successful a greater percentage of the time. And perhaps >those authors share their methods, and those who are watching take note and >learn. > >And again, the list is fundamentally a blitz list, which may evolve to include >std someday. But given the fact that most internet chess is blitz, and taking >into account the lack of controls which can adversly affect std ratings to a >greater degree than blitz, one has to pause and consider the intent of the whole >thing. And so I don't take it too seriously, and just do it for fun. >Will Your conclusions on: all searching 6 ply is the same as searching all 10 ply i take dead serious. I'm fanatically against this, and willing to contribute to do a huge experiment. I have volunteers at the icc club from who i don't doubt they want to cooperate. Can run at 3 or 4 computers at the same time to get a big number of games. all having same version same book and same settings. I'll give you the honor of making an official publication of this in the ICCA with the results as they are, so that some idiots that have big titles finally stop saying something that is against all logics and against the line of results we have seen. It's dead simple i think. If i go try to run the 100 metres right now, then i might need say 20 seconds. With some practice i might improve to say 13 seconds. With the same effort i can't improve another 7 seconds. With a lot of effort i even don't improve 7 seconds, i would run a big world record then! Vincent Diepeveen www.diepchess.com
This page took 0.01 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.