CLOP - what strategy?

[flok]

Hi,

I'm experimenting with CLOP.
It all seems to work and it seems to find parameters for improving the play of my engine.
But I'm not entirely sure if my method is the one to go.
What I do is: I let POS (not DBP, am testing the older non-brute-force-one) play games against a random mover as it has almost the same elo-rating.
My guess was that I should let it play against an engine that is somewhat stronger. Is that the right way to go? Because what would one do with an engine do that is stronger than all others?
Playing against itself would also be counterproductive I think because then you get an engine which is tuned to play against itself.

[jdart]

Even a basic alpha-beta program that can count material will out-perform a random mover. So if you are not out-performing that, then you are doing something wrong.

Yes, I think you should use a somewhat stronger engine as an opponent. Or several such engines. If your engine is weak you could try something like MSCP, although I think that has an incomplete Winboard protocol implementation.

--Jon

[flok]

Jon,

POS is not an alpha/beta engine. It searches only 3 or 4 plies deep but in fact that 4 plies searching is not really important for this engine.