PROGRAM ANALYSE MODE

[Catherine]

Hi all programmers, i don't know if it is possible, but i done a remark. when a program play for example a game of 50 moves/ 5 min. and after it analyse the same game with the same time, it see some error (i mean some best move that it not saw in playing mode). my question is: is it possible to use the analyse algorithm to play, in order to have a best quality of game? God bless you.
Catherine

[Rein Halbersma]

Hi all programmers, i don't know if it is possible, but i done a remark. when a program play for example a game of 50 moves/ 5 min. and after it analyse the same game with the same time, it see some error (i mean some best move that it not saw in playing mode). my question is: is it possible to use the analyse algorithm to play, in order to have a best quality of game? God bless you.
Catherine

Analysis mode has infinite thinking time. At some point a playing engine should stop thinking and make a move. So analysis mode cannot be an engine algorithm. To get high quality games, simply increase the thinking time until you are satisfied or until your patience has run out.

[Catherine]

Hi, i don't know a lot of thing in programmation, but i'm not agree with you, when you say that time of analyse mode if infinite. with Truus, Aurora Borealis or Dream... you can set the analyse time. so i think that it is not infinite, because the analyse take at the fixe time, no more no less. can you explain me that situation. i want just to know.
best regard
Catherine

[BertTuyt]

Catherine, I agree with Rein that normally the analysis mode should not reveal errors of the specific program when analyzing with the same time settings. It could be the contrary, as is the case with Damage. To check if the actual move is worse compared with the best move, the program injects the actual move at every new search-depth. By doing so one can calculate the real difference in move scores, as the PVS with minimal window search will only generate upper bounds for the other moves.

There is one exception. Truus uses a backwards tracking mechanism for the analysis. And with a clever hash-table management, this could yield new insights. This method however is not applicable in a ""normal" game situation, as the future moves are not yet given.

Bert

[Catherine]

Thank Bert for the details. i understand now. thank again.
Please, can we know Damage's evolution?

[BertTuyt]

Please, can we know Damage's evolution?

Hope I understand your question in the right way.

As mentioned in a recent post, I mainly focus this year on automatic learning of evaluation patterns.
There are some shortcomings in the Damage evaluation, which basically are "easy" to correct in a manual way, but I wanted to take a different approach this time.

With respect to the Damage GUI I'm integrating Database functionality, and Im also working on a open standard for a Database (.odd, .oddx) .
Will provide more information in the next weeks.

Last but not least, based on the encouraging results of the Horizon engine recently , I might publish an updated free version of Horizon this year.

Bert

[Catherine]

thank for all Bert, it is exactly, what i wanted to know. we are waiting for the uptade version. thank again for anwsering at all tyhe questions thank.