a) I think you need to show us more of the game. MCTS programs can make strange looking moves form the human point of view.

b) What are the computational resources you are using? (what are you running on). Some of these programs normally run on rather powerful machines, and pachi is one of those. You might want to ask the developers what sort of performance to expect on YOUR machine.

I'm not sure if it is the only problem, but it looks to me like pachi is not using the opening book. My version definitely play from the book with those moves by black.

You could try playing directly on the command line, which could give you some more feedback.

If you want to try this, you could launch pachi with the arguments you listed and use the following commands:

boardsize 19
play black q16
genmove white
play black q3
genmove white

Does pachi still play nonsense moves?

Add a showboard at the end, reading GTP is painful (or does pachi autoshow the board each time?)

_________________
Geek of all trades, master of none: the motto for my blog mostlymaths.net

It seems to autoshow after a genmove command.



It looks like it's playing the same opening as it does on my computer, which is one of the disadvantages of using an opening book, I suppose.

GoGui might work. It's what I use on my Linux installation.

Hi all,

Thank you for the info on installing Pachi on Goban.

Comparing the GTP command line output I found out that the strange opening moves could be the result of Goban NOT successfully loading the book.dat

Specifying the full path to book.dat in the argument works!

Unfortunately the patterns are automatically loaded by Pachi from it's working directory (try running pachi command line from outside its folder, then you see it is not loading the patterns).

p.s. I found this out trough GoGui, where I had to fill in the working directory for Pachi to make sensible opening moves.

Best,

Ruud