Author: Thomas Mayer
Date: 17:45:06 01/28/04
Go up one level in this thread
Hallo Christian, first -> it is good to have List out of the subject here... this is a basic question. >> Unfortunately, things are not that simple. >> Changing a few parameters in the evaluation module, adding a few tactical >> extensions to the search will easily produce a very different play. > Different ja, aber nicht viel viel stärker!!! the achieved improvement in strength does not play a big role... I think with around 10 lines of code I could make e.g. TSCP 100-200 Elos stronger... simply implementing nullmove... is this now a new program ? Am I in the position to claim that it is original ? You see that it is really not that easy. Of course you will not find similar 10 lines for Crafty - therefor it is to strong. To improve Crafty you might need to work hard, that is correct. > Die entscheidende Frage in der ganzen Diskussion ist wann man von einem Clone > spricht und wann von einer eigenen Schöpfung. Was darf "importiert" werden > was ist unzulässig (zB. HT-, TB-Technik, Zeitalgos, ... ). I start here very simple: NOTHING. Only exception is Nalimov's EGTB-Code - because he gave permission that everybody can use it (also the compression algorithm by Andrew Kadatch) when he is informed that he uses it. Let's take 10 programmers. Tell them how bitboards work and that they should write a move generator. You will find afterwards at least 10 different approaches -> maybe even 11 or more because some did try several approaches. So in simple words: Cut & Paste is a big No-No ! And this does not depend on which part or whatever... Or maybe you read the explanation how hashing can be done in "Schach am PC" -> I am sure that our 10 programmers will have 10 different approaches - again. Now show them before they implement it in say Mini-Max the Crafty Code... After you have explained how it works... If they are honest persons they might take a look in the code, but only when they are stuck somewhere... Again I am sure you will get 10 different approaches... ah, okay, maybe only 9 because one black sheep is always in the family... I am absolutely sure about this -> I have seen several open source code around - I haven't found any similarities except that they all play chess in them... okay, I must admit that I haven't looked very much in them -> as I said above, usually I try to learn and read as much as possible about a certain technique and then I try to implement it... when I get totally stuck I try to take a look in a open source to see if they have this technique also inside and how it works there... but I want to know as much as possible about the technique, with Cut & Paste I wouldn't understand it at all - so when you program it yourself it should be for you way easier to improve it... Or in other simple words: if the structure of your hashtable is exactly the same then used in Crafty I would already get suspicious - a) is it possible that two guys have here the exactly same ideas ? Because Hyatts structures are very search dependant, he stores there many interesting values that you only can really use when the structure of your search is somewhat similar... Overtaken those structures without understanding them might lead to the problem that you can't explain it... Also there was once the statement that Bitboard-Engines will look always very similar... I doubt this... I am very sure that between let's say Crafty, Patzer, IsiChess and Pharaon you will not find ANY single line that is the same... Just my two cents about such a statements... Greets, Thomas
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.