Question(s)/tips/discussion about RomiChess

[Martin Thoresen]

Hello, a new fan here!

I am using the newest RomiChess P3k, with pondering and learning enabled.

However, it is only updating the learn.dat file now and then, not after each game. Is this correct?

And also, the size seems to be constant at 274 kB. Can't remember if it started on 274kB or 0 kB, but it hasn't grown unlike some stuff when my eyes watches beautiful women.

[Bill Rogers]

As far as the size goes it is always rounded to the nearest "k" and in fact may take hundreds of new bytes before you will notice a change in its size. Lets say for instance that it is exactly 4096 byes long ie. 4k then you must add 1,000 more byes before it becomes 5k.
Bill

[Michael Sherwin]

Hi Martin,

I am always happy to have someone join the RomiChess fan club!

Bill is essentially correct, except that the learn file is grown 10,000 entries at a time, which is ~274 KB.

The interface that is used must send a 'result' command to Romi for there to be learning. This is done in all interfaces (AFAIK) for computer vs. computer games. However, Arena for example does not send a 'result' command for human vs. computer games and therefore there will be no learning in that case. For other interfaces, I do not know.

To inspect the learn file type at the command line 'pr 0' and you will see how many move records are stored in it. Follow the sib and chd records to inspect the learn file further.

[Martin Thoresen]

Michael,

Thanks!

I am using Chessbase interface, with a Wb2Uci adapter. Could this be the source of the "random" writing of the learningfile? I check the timestamp of the learning file and to make a guess I'd say it's updated (on average) after each third game or so.

Should it write the learning file no matter what the outcome of the game? Win, draw, loss?

Best Regards,
Martin

[Michael Sherwin]

Romi should write to the learn file after any result command is recieved, nomatter what the result is. But, I know nothing about your setup, so I hope that someone else can help.

In human vs. computer games the human must resign lost games for a result command to be sent by Winboard.

I use Arena 1.08 and it works well for computer vs. computer games.

[Martin Thoresen]

Romi should write to the learn file after any result command is recieved, nomatter what the result is. But, I know nothing about your setup, so I hope that someone else can help.

In human vs. computer games the human must resign lost games for a result command to be sent by Winboard.

I use Arena 1.08 and it works well for computer vs. computer games.

Thanks Michael! I use it only for engine vs engine play. For now.