Author: Robert Hyatt
Date: 12:55:16 07/27/04
Go up one level in this thread
On July 27, 2004 at 15:20:32, Ingo Bauer wrote: >Hi > >>>Of course I delete the learning for both engines!!! All I wanted to do is having >>>the same clean start for both engines. The engine that is having white would >>>have the possibility to learn something for its "black" game that the other >>>engine could not do when having black first. >> >>That logic is broken. If an engine plays black, it learns for "both sides". If >>an engine is white, the same thing holds. Disabling learning seems wrong, since >>it is a part of each engine, depending on how well it is implemented. Tuning >>bits of a program on or off on a whim seems somehow wrong unless the goal is >>_not_ to measure the strength of the entire "entity" but rather to measure the >>strength of a subset... > >I think this is quite "nit-picking" (right term? Sorry if too strong!). Most of chess is about "nit-picking". :) My point was this. When someone writes a chess engine, he devotes a part of the time to the search and extensions, a part of the time to the evaluation, a part of the time to the book selection code, a part of the time to book learning, a part of the time to position learning, a part of the time to parallel search, and so forth. We make those choices at the time we do the work. Now, when you turn off a feature I wrote, you are saying "that part of your time was thrown away and is not going to be worthwhile in this match because I am not going to use that part." What if I did position learning and my opponent chose not to. The time I spent on position learning you just threw away. The time he saved by not doing it was spent elsewhere in the engine, and you did not throw that away. That was my point. An engine is equal to the sum of all its parts. Not the some of a selected sub-set of its parts... > >1. It is been done for both engines! Suppose one doesn't do it very well, or doesn't do it at all? >2. Ok. if I take you definition Then I have to change my match to "I want to >check how the engines behave "out of the box"" Which is not a bad experiment. IE when you buy a new car, do you test it with one spark plug removed?? >3. I did some engine-test for some other programmers who direclty adviced my >either to disable learning for every engine (but to remove doublets except one) >or to delete learn files after on round of the match. IMHO you simply got bad advice. If someone spends the time/effort to develop learning code, it ought not be turned off just because someone suggests that it is the right way to test. See above for why. > >What do you think about this proposal: I finish the match with deleted >learn-files, and we will see what the result is. Then I delete the lernfiles for >both engines and do the whole match again without deleting lernvalues half way >through the match. There is only ONE thing YOU have to do. After the first match >is finished (around thursday evening) YOU have to give a guess about the outcome >of the second match! If you agree to that I will do another 70h of testing! > That is a better experiment. Learning on. Learning off. See how the results change if they do. The difference should be attributable to learning. >We will see if there is any considerable difference (which I doubt with only 40 >games) to the first match. I think this dispute is about nothing! > >Regards >Ingo Perhaps it is. Or perhaps not. But in the absense of data, assuming it is about "nothing" seems a bit superficial. If you want to measure the influence of learning on the Nunn positions, that makes sense. But to just turn it off "because" doesn't...
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.