Author: Bruce Moreland
Date: 16:07:59 06/03/01
Go up one level in this thread
On June 03, 2001 at 13:46:42, Gian-Carlo Pascutto wrote: >On June 03, 2001 at 13:16:42, Bruce Moreland wrote: > >>>Intresting. So you found a definite tactical strength increase? >>>Any numbers on WAC? >> >>WAC isn't a good suite once you've made your engine capable. If you run it at >>a minute per test, on decent hardware, you'll get at least 296 of them right. >>Reducing time doesn't change this much. I use the ECM suite because it is >>harder, so making the program faster or better results in more solutions. ECM >>has bugs, which I ignore, since even with them it's still the best tactical >>suite. > >I understand. But if I'm not mistaken ECM is several times as large >as WAC. For now I am using WAC as a first test. I get 293 with basic >Crafty on my machine at 5 sec per move. > >If I fiddle with it and I start getting a few less, thats a good >indication its not going to be tactically stronger with it. If you drill a hole in the side of it and grease starts leaking out, you will probably get less, but aside from doing that it would be hard to make it get any less. >My first attempts were pretty disasterous. I've done a lightweight >implementation based on code from Bob and your comments, and >now I get 292. Just one less, so it could be interesting. Five seconds isn't enough. You're probably wanting to tune searches longer than this. Even in a blitz program, you want at least ten seconds. Most of the solutions are a second or two, so it's hard to see any improvement. You don't just get the final number when you run a suite, you also get how long it took to solve each problem, and that's useful information that shouldn't be discarded. Using that suite to do anything real is just wrong. It's the wrong tool for just about anything. You might as well try to do eye surgery with a claw hammer and a box of toothpicks. >>I do something approximating what is mentioned in that paper. I don't mean to >>be a butthead, but I'm not going to go look at my stuff and lay it out here in >>pseudo-code. If I'm going to do that, I'd at least do it for a paper or >>something. > >I understand. > >>There are others who will take without crediting, and I don't want to >>be responsible for inflating their myth. > >I wonder if you're thinking of someone in particular :) No, I am not thinking of anyone in particular. I'm thinking of all of the professionals collectively. I tried to think of a nicer sounding way to put it, but that way works well enough. "Inflating their myth" sounds harsh but I don't mean it that way. >>I think that Bob started messing with it because I told him that I was messing >>with it and liked it. I am not sure if he had success with it or not. > >So far not I guess, as there's no sign of this code in standard crafty. > >>It's hard to get it to do anything other than suck plies. > >Agreed :) > >>I wrote a fairly direct implementation and it worked alright after some >>fiddling. I didn't do everything the DT guys did. I did it as cheaply as I >>could. > >I think this is indeed the right approach, as nowadays the programs >search much smaller trees, and hence something as costly as full SE >will just eat plies. If you lose more than 2 on every positions, >you can't ever hope to regain those with an extension. > >As a sidenote, do you have a copy of Louget Chess Test I lying around? >I can find II everywhere but I nowhere. I know you have used it for >testing, so it can't be hidden away too far? I use LCT1 as a dummy test, since it only takes 98 minutes to run. I use it to make sure I'm not wrecking middlegame or endgame performance. I use it to make sure I don't introduce bugs when I add a pure performance modification (one that doesn't change the size and shape of the tree). I don't tune to the positional positions or I'd drive myself insane. I have a tool that computes the scores from a dump file, but I don't use it very often. You can make big positive or negative changes without making the output score change at all. bruce > >-- >GCP
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.