Patricia 3

[pohl4711]

Mention, in the books for Shredder, Arena and Polyglot all moves have the normal probabilities (the more often the move is found in the book, the more often it is played).

Many thanks, I noticed (ctg and bin book in DroidFish) the probability is just evenly distributed, eg this is after e4:
Book randomisation is 0 and tourna/main lines off.

Yes, that is correct. Because the book is so small, in the FritzGUI/Droidfish-book all moves have the same probabilities. As far as I know, this is not possible in Arena/Shredder books, so there, I could not do it. So in these books (and polyglot, because it is a converted Arena-book) the moves have not the same probability to be played.

[Archimedes]

Some time ago I wrote a small batch file for the creation of Polyglot opening books. Unzip the archive into a directory of your choice. Simply copy the PGN files (several PGN files can be processed one after the other) into the input directory and start the batch file start.bat (the first two questions can be answered with no here). Among other things, you can also set the weights to 1 (all moves have the same probability of being played). The generated Polyglot opening books are then located in the output directory. For further information, see readme file.

PGNBook 2023.12.07:
https://app.box.com/s/kayeoj7lrjwotp96izrijz5p8joa0ory

[tapio]

Some time ago I wrote a small batch file

Thank you, will look at it. I recently used the Polyglot Book Factory in LucasChess, which makes things quite easy.

[tapio]

I cannot add the engine to LucasChess, shows cogwheel all the time. Is it maybe not compiled for older CPUs? I've got a Corei7-2600k, 2011 era.

[chessica]

I cannot add the engine to LucasChess, shows cogwheel all the time. Is it maybe not compiled for older CPUs? I've got a Corei7-2600k, 2011 era.

My PC is almost 20 years old and I was able to compile the source files. The engine works perfectly but is a clear step backwards compared to the 3.0 version. See here: viewtopic.php?p=967151#p967151

[tapio]

I have uploaded a set of 1-5 move polyglot books, to start into a game human vs. engine with some variety. Zip has a readme, I hope it makes sense. I'm by no means an expert in this.

[schack]

The evaluation is incredibly jumpy in multi-pv.

[Whiskers]

The evaluation is incredibly jumpy in multi-pv.

The eval is jumpy because of the aspiration windows being set; if you have 5 moves searched in multiPV mode, and the best one is +0.7 and the 5th best one is -4.5, a new depth iteration uses the windows used for the last move - i.e. really low windows that have to jump up a lot to catch up to the best move's score. I'm not 100% sure how to cleanly fix this at present. There's also an issue with reporting odd scores such as M500 because the evaluation function isn't properly clamped. I'll fix these little issues eventually.

[Sylwy]

The evaluation is incredibly jumpy in multi-pv.

Often erroneous:

[reflectionofpower]

Comparing its evaluation to another cookie cutter program will produce a different outcome. That's to be expected.