Winboard Forum

Hello Gentlemen !
Two questions-if possible:
A. Frequent I set the hash allocation in config.ini file or on various GUI depending of the chess engine type-but in strange mode -almost invariably I see in task manager a largest amount of used memory.Must I understand the hash amount in task manager shows the total allocation for the running .exe (transposition hash,pawn hash & so on) ?
B.I like to analyze the positional learning function of some chess programs (ProDeo,Hiarcs,Shredder,Spike,SmarThink.Gandalf and a very good one of The Baron) but I have a misunderstand: this function must be enabled in all situations ( engine vs. engine matches add-in) or mainly for analyse ?
Thank you a lot for your kindness !
Regards,
Silvian

Silvian Rucsandescu wrote:...Must I understand the hash amount in task manager shows the total allocation for the running .exe (transposition hash,pawn hash & so on) ?...

Yes it shows the total allocation. If you change the hash size it should change too.

Silvian Rucsandescu wrote:... this function must be enabled in all situations ( engine vs. engine matches add-in) or mainly for analyse ?
...

Sorry, I don't understand that question.

Volker

Silvian Rucsandescu wrote:Hello Gentlemen !
B.I like to analyze the positional learning function of some chess programs (ProDeo,Hiarcs,Shredder,Spike,SmarThink.Gandalf and a very good one of The Baron) but I have a misunderstand: this function must be enabled in all situations ( engine vs. engine matches add-in) or mainly for analyse ?
Thank you a lot for your kindness !
Regards,
Silvian

Spike has a very effective learning system. It saves positions on learning file when they have a spread of at least +- 20 centipawns from latest saved position. It updates learning files while playing or doing an analysis, so you haven't to play a game to the end to use its learning features. The Baron is a private engine since v. 1.8. Latest available version, 1.7.0, has not the very effective learning system you are talking about. If you want to have the latest beta (2.3), the only way is to ask to Richard Pijl, The Baron Author. I have it only because I'm a Baron tester.
Baron learning is derivated from Romichess. You can download Romichess here:

http://wbec-ridderkerk.nl/html/details/RomiChess.html

Romichess is far weaker of The Baron or Spike, since now. Mike Sherwin (the Author) is working hardly to improve the engine. From latest betas, I see good progresses. Romichess learning consists of applying a bonus score to moves which leaded to wins, a penalty score to moves which leaded to lost, and a very little penalty to moves which leaded to draw. As for The Baron, Romichess learning works only if game is over.

If you did read about last test-matches, both The Baron and Romichess obtained unexpected great results against Crafty and Glaurung. That was thanks to Manhattan.abk book, an opening book I built and structured to help learning features of Romi and Baron. If an engine uses it against another learning engine, the engine with the best learning system will eventually greatly improve its performance.
Manhattan book will be available from WBEC Romichess page soon. It has been sent to Leo about one week ago for uploading, but I guess he's very busy with tournament management now. So, he have just to wait for it....

Bye,