Arasan 14.0.1 (with UCI promotion fix)

[jdart]

Available from http://www.arasanchess.org.

--Jon

[Max]

Hi,

just tried to compile Arasan 14.0.1 on my old dusty PowerPC G4 Mac. All went well with gcc 4.2.1, but the created binary seems not to calculate normal chess moves.

Code: Select all

$ ./arasanx-32 
Arasan 14.0.1 Copyright 1994-2012 by Jon Dart. All Rights Reserved.
uci
id name Arasan 14.0.1
id author Jon Dart
option name Hash type spin default 32 min 4 max 1000
option name Ponder type check default true
option name MultiPV type spin default 1 min 1 max 10
option name OwnBook type check default true
option name Book selectivity type spin default 40 min 0 max 100
option name Threads type spin default 1 min 1 max 32
uciok
go
info nodes 22 hashfull 0
info depth 10 score cp 0 time 10 nodes 275 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 11 score cp 0 time 10 nodes 302 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 12 score cp 0 time 10 nodes 329 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 13 score cp 0 time 10 nodes 356 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 14 score cp 0 time 10 nodes 383 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 15 score cp 0 time 10 nodes 410 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 16 score cp 0 time 10 nodes 439 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 17 score cp 0 time 10 nodes 468 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 18 score cp 0 time 10 nodes 493 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 19 score cp 0 time 10 nodes 522 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 20 score cp 0 time 10 nodes 551 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 21 score cp 0 time 10 nodes 580 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 22 score cp 0 time 10 nodes 605 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 23 score cp 0 time 10 nodes 634 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 24 score cp 0 time 10 nodes 663 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 25 score cp 0 time 20 nodes 692 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 25 score cp 0 time 20 nodes 719 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
info depth 25 score cp 0 time 20 nodes 719 hashfull 0 pv a1a1 a1a1 pv a1a1 a1a1
bestmove a1a1 ponder a1a1a1a1
quit
$ 

The default compile options while make:

Code: Select all

...
g++-4.2 -O3 -fno-rtti -fno-enforce-eh-specs -fno-strict-aliasing -finline-functions    -D_MAC -Wall  -DUSE_INTRINSICS -DUSE_ASM  -DSMP -DSMP_STATS  -c -o ../build/arasanx.o arasanx.cpp
...  

Anything to change in the provided Makefile?

All the best,
Max

[jdart]

From the programr.txt file in the doc directory:

"Arasan has only been tested on Intel & AMD processors. The code is not
written to be portable to other processors. If you want to run on
non-x86 hardware, you will likely have to make source
modifications. In particular, there is only limited support (not fully
tested) for big-endian architectures in the source."

--Jon

[Max]

Thanks Jon,

for your well written documentation ... and this ' ' didn't read it.

Sorry,
Max

[lucasart]

Available from http://www.arasanchess.org.

--Jon

Arasan seems to lose on time quite often. For example, after 500 games in 1'+1", with a tolerance of 1", using cutechess-cli (whose "lag" is certainly below 0.01"), I've had 5 time losses:

Code: Select all

lucas@megatron:~/Chess/Games$ grep forfeit gauntlet_arasan_14.0.pgn 
[Termination "time forfeit"]
[Termination "time forfeit"]
[Termination "time forfeit"]
[Termination "time forfeit"]
[Termination "time forfeit"]

The games are typically long ones, that would otherwise have ended in a 50 move draw. Note that I never adjudicate any game, and let them finish by chess rules (mate, 3/50 draw, stalemate). Given 1" increment and 1" tolerance, there should never be any time loss.

Perhaps you're trying to do too complex things in your time management code, and it ends up backfiring in certain circumstances. In particular, you should always make sure that you don't allocate more than the remaining time plus a safety margin. Here's what I do in DoubleCheck, to give you an idea:

Code: Select all

min(blitz_time/40 + blitz_inc, blitz_time-25)

* blitz_time is the remaining time (in ms)
* blitz_inc is the increment (in ms)
So I basically allocate blitz_time/40+blitz_inc, but I cap this value to blitz_time-25 to be on the safe side. Of course my time code is very simplistic and in no way optimal, but it's a defensive approach that proved to work (no time losses ever in 6"+0.1" for example).

[IGarcia]

Arasan seems to lose on time quite often. [...]

I observed the same, but since i was using a beta GUI was unsure to post the problem. In a first match, when used Arasan 14.0 there was only one time forfit in a position with some disadvantage for Arasan. Match was discarded because later the promotion bug appeared, and started a new one. Now using Arasan 14.1, against Gaviota 0.85, 60 games, 5+3" and got 13 time forfeits, 10 of them in lost position and 3 in a draw position.

Some observation:

One reason to discard the first match, besides the promotion bug, was a misconfiguration of my part. Arasan played the match using Nalimov TB witch I have incomplete. Only 4 men + a few 5 men. So was not fair. In the second match, I configured Arasan to use Gaviota TB. And by checking the 13 troubled games, there where tablebase hits in the final position of all of them. So, 60 games using Nalimov got 1 time forfeit versus 60 games with 13 time forfeits when using Gaviota maybe is a clue.

Regards,
Ignacio.

[stevenaaus]

On my 32bit linux, it is broke with ponder on. It gives malformed ponder move, and then hangs.
Engine: bestmove b1c3 ponder b1c3b8c6

Ponder off seems fine.

md5sum arasan-14.0.1.tar.gz - 2ef56017e543ff04c090dea3ded52f82

Engine: Arasan 14.0.1 (broke)
Command: arasanx-32
Date: Sun Mar 11 12:11:36 EST 2012

This file was automatically generated by Scid.
It is rewritten every time an engine is started in Scid.

NOTE : First line from engine seen; sending it initial commands now.
Scid : uci
Engine: Arasan 14.0.1 Copyright 1994-2012 by Jon Dart. All Rights Reserved.
Engine: id name Arasan 14.0.1
Engine: id author Jon Dart
Engine: option name Hash type spin default 64 min 4 max 1000
Engine: option name Ponder type check default true
Engine: option name MultiPV type spin default 1 min 1 max 10
Engine: option name OwnBook type check default true
Engine: option name Book selectivity type spin default 40 min 0 max 100
Engine: option name Threads type spin default 1 min 1 max 32
Engine: uciok
Scid : isready
Engine: Using book file book.bin
Engine: readyok
Scid : setoption name Hash value 64
Scid : isready
Engine: readyok
Scid : setoption name MultiPV value 1
Scid : isready
Engine: readyok
Scid : setoption name OwnBook value false
Scid : isready
Engine: readyok
Scid : setoption name Book selectivity value 40
Scid : isready
Engine: readyok
Scid : setoption name Threads value 1
Scid : stop
Scid : ucinewgame
Scid : setoption name Ponder value true
Scid : isready
Engine: readyok
Scid : position startpos
Scid : go wtime 6000 btime 6000 winc 0 binc 0
Engine: info nodes 22 hashfull 0
Engine: info depth 5 score cp 28 time 10 nodes 3587 hashfull 0 pv d2d4 d7d5 g1f3 c8f5 c1f4 pv d2d4 d7d5 g1f3 c8f5 c1f4
Engine: info depth 6 score cp -4 time 10 nodes 5106 hashfull 0 pv d2d4 d7d5 g1f3 c8f5 c1f4 g8f6 pv d2d4 d7d5 g1f3 c8f5 c1f4 g8f6
Engine: info depth 6 score cp 0 time 10 nodes 6441 hashfull 0 pv b1c3 b8c6 e2e3 e7e5 f1c4 d8f6 pv b1c3 b8c6 e2e3 e7e5 f1c4 d8f6
Engine: info depth 7 score cp 24 time 40 nodes 17783 hashfull 1 pv b1c3 e7e6 g1f3 f8b4 e2e4 g8f6 f1d3 pv b1c3 e7e6 g1f3 f8b4 e2e4 g8f6 f1d3
Engine: info depth 7 score cp 28 time 40 nodes 20205 hashfull 1 pv d2d4 d7d5 g1f3 c8f5 c1f4 g8f6 b1c3 pv d2d4 d7d5 g1f3 c8f5 c1f4 g8f6 b1c3
Engine: info depth 8 score cp -4 time 60 nodes 27902 hashfull 1 pv d2d4 d7d5 g1f3 g8f6 c1f4 c8f5 b1c3 b8c6 pv d2d4 d7d5 g1f3 g8f6 c1f4 c8f5 b1c3 b8c6
Engine: info depth 8 score cp 12 time 70 nodes 34277 hashfull 2 pv b1c3 e7e6 d2d4 g8f6 c1f4 b8c6 e2e4 f8b4 pv b1c3 e7e6 d2d4 g8f6 c1f4 b8c6 e2e4 f8b4
Engine: info depth 9 score cp 4 time 100 nodes 52147 hashfull 3 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1
Engine: info depth 10 score cp -33 time 160 nodes 80049 hashfull 4 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1
Engine: info depth 10 score cp -33 time 160 nodes 80049 hashfull 4 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1 pv b1c3 b8c6 e2e3 g8f6 g1f3 d7d5 f1b5 c8d7 e1g1
Engine: bestmove b1c3 ponder b1c3b8c6
Scid : position startpos moves b1c3 b1c3b8c6
Scid : go ponder wtime 5827 btime 6000 winc 0 binc 0
Scid : stop
Scid : position startpos moves b1c3 b8c6
Scid : go wtime 5827 btime 5647 winc 0 binc 0
Engine: info depth 1 score cp -947 time 10 nodes 669 hashfull 4 pv g1f3 c3d1 e1d1 pv g1f3 c3d1 e1d1
Engine: info depth 1 score cp 52 time 10 nodes 719 hashfull 4 pv d2c3 e7e5 pv d2c3 e7e5
Engine: info depth 2 score cp 15 time 10 nodes 743 hashfull 4 pv d2c3 e7e5 pv d2c3 e7e5
Engine: info depth 2 score cp 8 time 10 nodes 765 hashfull 4 pv d2c3 d7d5 pv d2c3 d7d5
Engine: info depth 3 score cp 45 time 10 nodes 827 hashfull 4 pv d2c3 e7e5 pv d2c3 e7e5
Engine: info score cp 45 lowerbound
Engine: info depth 3 currmove d2c3 currmovenumber 1 nodes 827 hashfull 4
Engine: info depth 3 score cp 48 time 10 nodes 947 hashfull 4 pv d2c3 e7e5 g1f3 pv d2c3 e7e5 g1f3
Engine: info depth 4 score cp 11 time 10 nodes 1124 hashfull 4 pv d2c3 e7e5 g1f3 pv d2c3 e7e5 g1f3
Engine: info depth 4 score cp 8 time 10 nodes 1236 hashfull 4 pv d2c3 e7e5 g1f3 d8f6 pv d2c3 e7e5 g1f3 d8f6
Engine: info depth 5 score cp 45 time 10 nodes 2023 hashfull 4 pv d2c3 e7e5 pv d2c3 e7e5
Engine: info score cp 45 lowerbound
Engine: info depth 5 currmove d2c3 currmovenumber 1 nodes 2023 hashfull 4
Engine: info depth 5 score cp 52 time 10 nodes 3062 hashfull 4 pv d2c3 d7d5 c1f4 c8f5 g1f3 pv d2c3 d7d5 c1f4 c8f5 g1f3
Engine: info depth 6 score cp 16 time 10 nodes 3900 hashfull 4 pv d2c3 d7d5 c1f4 c8f5 g1f3 g8f6 pv d2c3 d7d5 c1f4 c8f5 g1f3 g8f6
Engine: info depth 7 score cp 28 time 10 nodes 5281 hashfull 4 pv d2c3 d7d5 c1f4 c8f5 g1f3 g8f6 e2e3 pv d2c3 d7d5 c1f4 c8f5 g1f3 g8f6 e2e3
Engine: info depth 8 score cp 16 time 20 nodes 8928 hashfull 5 pv d2c3 d7d5 c1f4 c8f5 g1f3 b8c6 f3d4 c6d4 c3d4 pv d2c3 d7d5 c1f4 c8f5 g1f3 b8c6 f3d4 c6d4 c3d4
Engine: info depth 9 score cp 16 time 30 nodes 13321 hashfull 5 pv d2c3 d7d5 c1f4 c8f5 g1f3 b8c6 f3d4 c6d4 c3d4 pv d2c3 d7d5 c1f4 c8f5 g1f3 b8c6 f3d4 c6d4 c3d4
Engine: info depth 10 score cp 12 time 70 nodes 32698 hashfull 6 pv d2c3 d7d5 g1f3 b8c6 c1f4 c8f5 f3h4 e7e5 h4f5 e5f4 pv d2c3 d7d5 g1f3 b8c6 c1f4 c8f5 f3h4 e7e5 h4f5 e5f4

[jdart]

I'll take a look at the ponder issue, thanks for the info.

--Jon

[jdart]

Are you running the compiled linux binary I built or did you build it on your own from source?

--Jon

[stevenaaus]

It occurs with both your 32 bit exe and my compiles.