Best Fischer personality?

[David Dahlem]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

[Dr.Wael Deeb]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

Pretty much the answer is no....

[Ovyron]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

Chessmaster's will ensure that the opening will be played like Fischer as they include a book specialized for this task. Otherwise, Vas said that this is just nonsense.

[David Dahlem]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

Chessmaster's will ensure that the opening will be played like Fischer as they include a book specialized for this task. Otherwise, Vas said that this is just nonsense.

Hey, i just had an idea! RomiChess has a learning option, and it's possible to merge a pgn file into the RomiChess learn file. Perhaps after merging a pgn of all Fischer games, then write protecting the learn file, RomiChess will play like Bobby Fischer?



Regards
Dave

[Michael Sherwin]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

Chessmaster's will ensure that the opening will be played like Fischer as they include a book specialized for this task. Otherwise, Vas said that this is just nonsense.

Hey, i just had an idea! RomiChess has a learning option, and it's possible to merge a pgn file into the RomiChess learn file. Perhaps after merging a pgn of all Fischer games, then write protecting the learn file, RomiChess will play like Bobby Fischer?



Regards
Dave

There is one more undocumented thing that Romi can do.

First, use a strong tactical engine like Rybka or Shreddar and proof all of Fishers games one at a time so that all of his games are truncated at the first Fischer blunder. Mark all the games as Fischer wins. Even his draws and loses.

Then start Romi at the command line with a new learn file.

Type 'book 2'

Type 'merge fischer.pgn'

Type 'quit'

Now Romi will mark as permanent book moves all of Fischer's moves and Randomly pick between them. Romi will still use her learning once out of the permanent Fischer book.

Edit: However, Vas is right that no program can simulate a particular humans play.

[David Dahlem]

To the best of my knowledge, there are three engines with Fischer personalities: ProDeo, Chessmaster, and DeepSjeng. Which these personalities play more like the Fischer style (whatever that is)?

Regards
Dave

Chessmaster's will ensure that the opening will be played like Fischer as they include a book specialized for this task. Otherwise, Vas said that this is just nonsense.

Hey, i just had an idea! RomiChess has a learning option, and it's possible to merge a pgn file into the RomiChess learn file. Perhaps after merging a pgn of all Fischer games, then write protecting the learn file, RomiChess will play like Bobby Fischer?



Regards
Dave

There is one more undocumented thing that Romi can do.

First, use a strong tactical engine like Rybka or Shreddar and proof all of Fishers games one at a time so that all of his games are truncated at the first Fischer blunder. Mark all the games as Fischer wins. Even his draws and loses.

Then start Romi at the command line with a new learn file.

Type 'book 2'

Type 'merge fischer.pgn'

Type 'quit'

Now Romi will mark as permanent book moves all of Fischer's moves and Randomly pick between them. Romi will still use her learning once out of the permanent Fischer book.

Edit: However, Vas is right that no program can simulate a particular humans play.

So, if i have a large collection of Rybka (or any engine or human) games, and i disregard any possible blunders, and i mark all that players games as wins, Romi will randomly pick from that players moves up to ply 60?

That's a great feature, if i'm understanding correctly.

Regards
Dave

[Philippe]

Michael,
You should put all these undocumented features in the readme.text file.

[David Dahlem]

Another idea, suppose you took a large collection of GM games, for example enormous.pgn from Hyatt's site, removed all draws, separate white wins and black wins into two pgn files, then used the global search and replace feature of your text editor to change the name of the winning player in both files to a common ficticous name, such as "Winner", merge these two pgn files into one, then start Romi at the command line with a new learn file, then

Type 'book 2'
Type 'merge Winner.pgn'
Type 'quit'

Seems like that would create a very strong learn file/opening book, right?

Regards
Dave

[Michael Sherwin]

Michael,
You should put all these undocumented features in the readme.text file.

Sorry, you are right ofcourse. However, I am just a pinball in a fast and furious game at the moment and I did not manage to hit that particular bumper.

[Michael Sherwin]

Another idea, suppose you took a large collection of GM games, for example enormous.pgn from Hyatt's site, removed all draws, separate white wins and black wins into two pgn files, then used the global search and replace feature of your text editor to change the name of the winning player in both files to a common ficticous name, such as "Winner", merge these two pgn files into one, then start Romi at the command line with a new learn file, then

Type 'book 2'
Type 'merge Winner.pgn'
Type 'quit'

Seems like that would create a very strong learn file/opening book, right?

Regards
Dave

If the games are not proofed, book 2, will produce a very bad book. As even the winning side makes blunders (computer games between strong engines might be doable). Just using merge with out book 2 will produce a better book, because the moves will be chosen statistically rather than randomly.