UCI Engine Tuning

[hgm]

And fairymax leaves out the promotion piece altogether ? ... Bad.

The version I have (4.8 R) does print a promotion piecein its move. Just not in its PV. (Always 'q' in normal Chess, as Fairy-Max' search does not implement under-promotion. It does accept under-promotions on input, though.)

[stevenaaus]

And fairymax leaves out the promotion piece altogether ? ... Bad.

The version I have (4.8 R) does print a promotion piecein its move. Just not in its PV. (Always 'q' in normal Chess, as Fairy-Max' search does not implement under-promotion. It does accept under-promotions on input, though.)

Is this version available ?

I've done some work on the uci and xboard code. Xboard seems to have fundamental differences between playing white (send "go") and black (send "new"). And the UCI doco i have mentions "debug" or "debug off" - but these seem unsupported by at least a few engines, and even "ucinewgame" isnt supported sometimes. .....

[hgm]

A "new"command resets a WB engine to a state where it plays back. That is good if you want it to play black, but if you want it to play white, you need an extra "go" to change that state, and set it thinking where it would otherwise ust be waiting for the opponent's input move.

The latest Fairy-Max is available from my on-line repository where you got the XBoard source ( http://hgm.nubati.net/cgi-bin/gitweb.cgi ). Under Linux a simple "sudo make install" should do, there is no config.

[stevenaaus]

Xboard support seems better, including non-standard start positions. Cheers.
I omitted to test Phalanx... too hopeless ;>

Code: Select all

------------------------------------
 1: Arasan131   3.5 /  4   XX =1 11
 2: Gnu Chess   1.5 /  4   =0 XX 01
 3: Fairy       1.0 /  4   00 10 XX
------------------------------------

Two more things... I've added a "Copy" button to the Scid vs. PC's engine configuration widget, so it's easy to test an engine against itself with different UCI configurations. The only caveat is - I no longer use the engine name supplied by "id name" for the computer tournament.. because there's no telling the two engines apart - funnily enough

And I did *some* work trying to reduce the interface overhead, but .... hmmm .... it's never going to beat a dedicated C tournament GUI , though Scid does have a lot of other things going for it.