Skip to content
Permalink
Browse files

sf_10 without adjudication

sf_10 without adjudication disables with fake scores sent.
Bench:3939338
  • Loading branch information...
Chess13234 committed Dec 27, 2018
1 parent b4c239b commit 449497a5a9241be1991508b8f89e807cc3bfe100
Showing with 2 additions and 5 deletions.
  1. +2 −5 src/uci.cpp
@@ -256,11 +256,8 @@ string UCI::value(Value v) {

stringstream ss;

if (abs(v) < VALUE_MATE - MAX_PLY)
ss << "cp " << v * 100 / PawnValueEg;
else
ss << "mate " << (v > 0 ? VALUE_MATE - v + 1 : -VALUE_MATE - v) / 2;

v =PawnValueEg;
ss << "cp " << v;
return ss.str();
}

14 comments on commit 449497a

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@vdbergh it will be a separate test. One moment.

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@vondele

This comment has been minimized.

Copy link

replied Dec 27, 2018

@vdbergh @Chess13234 ... isn't the game adjudicated by cutechess only when both engines agree on a score (I'm not 100% certain) ? Therefore it is enough if one of the two engines are reporting fake scores. As such, there are now two identical tests running AFAICT.

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@vondele Not. Only draws are adjudicated by both.

-draw movenumber=NUMBER movecount=COUNT score=SCORE
Adjudicate the game as a draw if the score of both
engines is within SCORE centipawns from zero for at
least COUNT consecutive moves, and at least NUMBER full
moves have been played. Captures and pawn moves will
reset the counters.
-resign movecount=COUNT score=SCORE
Adjudicate the game as a loss if an engine's score is
at least SCORE centipawns below zero for at least COUNT
consecutive moves.

@vdbergh

This comment has been minimized.

Copy link

replied Dec 27, 2018

@vdbergh @Chess13234 ... isn't the game adjudicated by cutechess only when both engines agree on a score (I'm not 100% certain) ? Therefore it is enough if one of the two engines are reporting fake scores. As such, there are now two identical tests running AFAICT.

@vondele Good point but this is what https://manpages.debian.org/testing/cutechess/cutechess-cli.6.en.html says.

-resign <n> <score>
Adjudicate the game as a loss if an engine’s score is at least <score> centipawns below 
zero for at least <n> consecutive moves.

This seems to suggest that it is not necessary for both engines to agree.

@joergoster

This comment has been minimized.

Copy link

replied Dec 27, 2018

Afaik, cutechess does recognize the score of both engines even with resign option set. Yet only one side (the losing one) can resign, of course!
At least, I've never seen a single game where one side resigned and the other engine showing a significantly different score, resp. a not winning score.

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@joergoster That "winning score" could be wrong.

@joergoster

This comment has been minimized.

Copy link

replied Dec 27, 2018

@Chess13234 Then show some games where this is the case, please.

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@joergoster Check the linked fishcooking discussion threads in the main issue, i think they already found some in the past.
official-stockfish#1904

@Chess13234

This comment has been minimized.

Copy link
Owner Author

replied Dec 27, 2018

@vdbergh strangely that version doesn't mention draw counters being reset.(i use cutechess compiled from last weeks git and it does includes "Captures and pawn moves will
reset the counters.").
-draw <n> <score> Adjudicate the game as a draw if the score of both engines is within <score> centipawns from zero after <n> full moves have been played.

@joergoster

This comment has been minimized.

Copy link

replied Dec 27, 2018

@vondele @vdbergh It looks like cutechess only takes the score from one side to make a resign decision if I read that code correctly.
https://github.com/cutechess/cutechess/blob/master/projects/lib/src/gameadjudicator.cpp#L105-L120

@Chess13234 Admittedly, this seems a bit unfortunate.

@vdbergh

This comment has been minimized.

Copy link

replied Dec 28, 2018

I would suggest increasing the throughput of this test to 500. There seems to be little point in making it run forever.

@vondele

This comment has been minimized.

Copy link

replied Dec 28, 2018

done, but generally, these tests should be done sparingly.

@vdbergh

This comment has been minimized.

Copy link

replied Dec 28, 2018

I agree they should done sparingly. But once they are running there seems to be no point in artificially delaying them.

Please sign in to comment.
You can’t perform that action at this time.