Author: Marc van Hal
Date: 04:24:59 06/13/02
Go up one level in this thread
On June 13, 2002 at 00:19:44, Uri Blass wrote: >On June 12, 2002 at 19:56:49, Marc van Hal wrote: > >>On June 12, 2002 at 12:26:33, Uri Blass wrote: >> >>>On June 12, 2002 at 12:00:37, Marc van Hal wrote: >>> >>>>On June 11, 2002 at 12:36:41, Christophe Theron wrote: >>>> >>>>>On June 11, 2002 at 01:42:10, Uri Blass wrote: >>>>> >>>>>>On June 10, 2002 at 22:29:44, Christophe Theron wrote: >>>>>> >>>>>>>On June 10, 2002 at 15:28:14, Rajen Gupta wrote: >>>>>>> >>>>>>>>i have read somewhere (i think it was hinted in one of the interviews which >>>>>>>>frank morsch gave to one of the indian newspapers)that at any given time, there >>>>>>>>are several different versions of fritz being developed:- the inference being >>>>>>>>that and the one that is actually released is not necessarily the strongest one; >>>>>>>>its the one that is just strong enough. >>>>>>>> >>>>>>>>frank morsch apparently has one ready whenever a new upstart arrives on the >>>>>>>>scene.i wont be surprised if there is no new fritz till something overtakes the >>>>>>>>current version. >>>>>>>> >>>>>>>>rajen >>>>>>> >>>>>>> >>>>>>> >>>>>>>It does not make sense. >>>>>>> >>>>>>>Look at the small margin between Fritz and the program just behind it (Tiger) on >>>>>>>the SSDF. >>>>>>> >>>>>>>Why would Frans take the risk of publishing an engine that might fail to achieve >>>>>>>the first place on the SSDF if he has something better? >>>>>> >>>>>>Maybe he does not know which engine is the best. >>>>>> >>>>>>The only way to be sure that engine A is better than engine B is by games. >>>>>>You can always have other tests in order to guess but they are only an estimate. >>>>>> >>>>>>I know that you say that you do not use games against other opponents but I >>>>>>think that it is a mistake. >>>>>> >>>>>>The fact that you probably have some test that usually gives >>>>>>the same results as games is a good reason to use that test for testing one >>>>>>change but when you decide to release a new version the only way to be sure that >>>>>>it is better is by a lot of games(unless the change is only doing tiger faster). >>>>>> >>>>>>Uri >>>>> >>>>> >>>>> >>>>>In order to have a top chess program you must have a method to decide if a >>>>>change is an improvement or not. One of the requirements of this method is that >>>>>you must be able to get a result in a short period of time (preferably less than >>>>>4 days in the most difficult cases). >>>>> >>>>>There are many little changes to test before you get a version significantly >>>>>stronger than your last release. >>>>> >>>>>It is not practical to let people test several versions and decide for you >>>>>because you can't rely on results you have not controlled yourself (there are >>>>>too many possibilities of inconsistencies even in the experiments you set up >>>>>yourself) and because these people would have to play a lot of games under >>>>>equivalent conditions in order to get statistical relevance (which you seldomly >>>>>get, because you cannot ask people to play 500 games in a row). >>>>> >>>>>I cannot believe that a serious chess programmer would use such a lousy >>>>>selection method. >>>>> >>>>>Testers feedback is very valuable to spot problems or lacks in the program's >>>>>knowledge, bugs, and more generally good advices on general directions to work >>>>>on. >>>>> >>>>>Testers feedback is used to get quality data, human advice and creativity, you >>>>>generally cannot use it to get a quantity of statistically relevant data. >>>>> >>>>>The final decision about what is an improvement and what is not must be taken by >>>>>the programmer himself, with a cold, scientifically controlled, objective, test >>>>>method. >>>>> >>>>> >>>>> >>>>> Christophe >>>> >>>>Chess is not science yet! >>>>You have some difeculty's from both sides >>>> >>>>a) the programmer does not have much chess knowledge. >>>>And he is the one who makes the decisions about the changes. >>>> >>> >>>I think that having much chess knowledge is not the important thing for chess >>>programmers. >>> >>>It is more important to know to explain the knowledge that you have to the >>>computer. >>> >>>I think that no programmer is close to be perfect in that task. >>>The important thing is to do the right observations and explain them to the >>>computer and not to have a lot of chess knowledge. >> >> >>Marc van Hal): For making the right observation you need to have chess >>knowledge! >>> >>>I also think that chess knowledge is something that >>>you need to find defintions for it by yourself. >> >> >>Marc van Hal): >>Ok Uri you made me thinking here and actualy i came to the conclusion we where >>all wrong! >>It might come as a shock for Amir (Sory for that Amir.) >>Because the best scienetificly explanition in human terms chess knowledge is >>nothing more or less then what you know from a position. >>based on other games played with that same kind of position! >> >>So in fact chessprogram don't have real chess knowledge! >> >>Or it must have created this knowledge by stored learning files! >> >> >> >>The programing of what we also call chess knowledge >>But in fact is helping to get a more precise evaluation of a position. >> >>(Which also is called evalution of a position in human terms only chess >>knowledge and the general chess rules these could be revieuwed perhaps >>by adding some rules on how to keep the tempo's. >>(Though Eric Schillers piece of the Siclian was very close to perfection.) >>does help human players to evaluate their positions!) >> >>is a very wide selection,of variables. >> >>(Some of these varables are not even used yet. >>pointing at the learning methodes of an engine. >> >>(Which in fact is the only way to increase real chess knowledge!) >> >>I explained a way it could be improved some weeks ago with my A B C search >>methode. >>Not to metion that I in the past pointed out that if a program finds a move >>which gets an increased evalution most of the time is beter then the move with >>the higher evalution. > >I know it. >I thought independently about the same thing(I did not read what you mentioned >in the past) but the problem is that programs do not evaluate the second best >move so they cannot know if the second best move has an increasing evaluation. > >Evaluating the second best move only to find if it has an increasing evaluation >is going to do the program significantly slower and the move with increasing >evaluation can be the 3th or the 4th best move. Marc van Hal): or even 5th or 6th I have analysed some kings Indian position with Junior7 and it did not show the right moves it only played this move at once after I played it. Marc van Hal):You could over ride this problem by before releasing the program play many games with a slower time control against much diferent oponents or by make it analyzing games with this slower time control. So at the moment an end user start to use it the learning file will speed up the search again. You also could make a diferent version of the program and insert the learn files in the normal version. > >I believe that I have better ideas that are used in Movei but they can help >movei only in cases of a small difference in the static evaluation(the >evaluation of a position is not dependent only on the final position but the >change that I allow in the static evaluation is only a small change so if the >best move is 0.2 pawns better than the second best(based on the static >evaluation that movei does not use) then there is no chance that movei is going >to choose the second best move. > >Note that you do not need a special chess knowledge to make that observation and >this is the reason that I say that the important thing is to make the right >observations. > I>Uri Marc van Hal): I don't know what you mean by special chess knowledge. I only can say that chessknowledge can help to find out what the program evaluates wrong,based upon this knowledge! and how to change the parameters so it will evalute the positions better. The beter the program the more knowledge is required to find this out. Chess programers try to find this out by making diferent version which is much more time consuming. Not to mention that if you only test these settings against other programs you never can get a real improvement. Meaning it will still make the same mistakes as the programs of that curent time. And you still are not sure if this are the best settings and if this will solve the evaluation problems it had with previous versions. Many times one evaluation problem is corected and an other evaluation problem pops up. And after inserting what we used to call chess knowledge All parameters should be revieuwed again to use this knowledge in the most optimal way!
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.