Matthias Gemuh wrote:Ferdy wrote:Matthias Gemuh wrote:hgm wrote:Matthias Gemuh wrote:1) Solving testsuites
- Some UCI engines treat "go infinite" in a strange manner, so solving
testsuites fails for them (Fire4, Gull3, etc).
Can you elaborate a bit on that?
They analyze FENs without crashing, but spit out PVs which do not start with the expected moves.
Why not use the bestmove engine output and compare it with the bm in epd test suite?
I tried Fire 4 with my tool and the first move in the pv is the bestmove sent.
ChessGUI does not unload the engines between FENs, so it does not send "isready" between FENs.
That tool that I run also does not quit the engine after receiving the engine bestmove in every position.
I just send the isready command and wait for the readyok reply just to make sure that the engine is there,
and it is ready before I send the ucinewgame followed by position and go commands.
I will not send the ucinewgame/position/go commands until I receive that "readyok".
Code: Select all
2015-08-09T21:14:59.732000 >> isready
2015-08-09T21:14:59.951000 << readyok
2015-08-09T21:14:59.951000 >> ucinewgame
2015-08-09T21:14:59.951000 >> position fen r3k1nr/pp1n2pp/1bp1N1q1/6B1/2B1p3/8/PPP2RPP/R2Q3K w kq - 0 1
2015-08-09T21:14:59.951000 >> go infinite
Matthias Gemuh wrote: Fire4 and Gull3 seem to expect that "isready".
BTW, bestmove approach may be added later, but Fire4 and Gull3 would still not benefit from "Fast Mode" option of ChessGUI.
In "Fast Mode", a position is considered solved if the first move of PV is accurate for X consecutive plys, or if the first move of PV is accurate after at least Y amount of time.
I tested Fire4 and Gull3 only in "Fast Mode".
It is good to know regarding that fast mode.
I see it when I tried ChessGUI. I was surprised when Deuterium could
only score 2/300 in WAC, even if the eval is showing mate in 2
or 3 or 4 or 5 or 6 or 7. I will experiment a bit more.