bright 0.4a now public

[BBauer]

In my first example I used bitbases.
So it looks like bright works as exspectet +without+ bitbases.
Bright +with+ bitbases may cause errors.
kind regards
Bernhard

[royb]

Hi everybody,

I decided to make version 0.4a public, it can be downloaded from:

http://members.ziggo.nl/allard.siemelink/bright/

major changes include search and (king safety) evaluation,
resulting in more attractive play and a ~50 elo gain compared
to version 0.3a, the previous public version.

Is there a chance you could make a 64-bit native Linux binary? I run 64-bit Ubuntu and would prefer to run a 64-bit version of Bright rather than the 32-bit Windows version with wine.

Thanks!

Roy

[Allard Siemelink]

Unfortunately, it is not entirely trivial to create a native linux binary as Bright calls the windows api to manage its threads.
Since Bright is not a bitboard engine I do not expect any speed gains either.

However, as I do run ubuntu 64 occaisionally, it is on my todo list, just do not expect it anytime soon.

Is there a chance you could make a 64-bit native Linux binary? I run 64-bit Ubuntu and would prefer to run a 64-bit version of Bright rather than the 32-bit Windows version with wine.

Thanks!

Roy

[Allard Siemelink]

That's odd. On my machine bright 0.4a finds the correct move at depth 2, both with and without bitbases.
Did you perhaps recently upgrade to a newer version of the bitbases? (I used egbbprobe 3.1)
Does it find the mate when you turn off the bitbases?

Not for me. I get

FEN: 2b3rk/5P2/7K/5N2/8/8/8/8 w - - 0 1

Bright-0.4a:
6/8 00:00 532 33 0,00 f7xg8Q+
7/8 00:00 967 60 0,00 f7xg8Q+
8/8 00:00 1.512 47 0,00 f7xg8Q+
9/10 00:00 2.284 71 0,00 f7xg8Q+

and bright computes for more than 2 min without any output.
kind regards
Bernhard

[Spock]

Naturally I'd like to fix all FRC bugs, could you perhaps e-mail me any problem games and/or logs you might have?

I've played 200 FRC games now under ChessGUI and it is running perfectly. Not sure what the earlier problem may have been last year. Hopefully it will continue to be fine

[Allard Siemelink]

Although I do not believe that we can conclude too much from a mere 18 games, I must admit that I too noticed that Bright seems to have trouble playing Thinker.

according to ccrl 40/4, the 4 cpu versions perform relatively well:

Thinker 5.4c Inert 64-bit 4CPU (+45) 44 − 56 (+26−38=36) 44.0%
Thinker 5.4a Inert 64-bit 4CPU (+20) 51.5 − 48.5 (+34−31=35) 51.5%
Thinker 5.3b Inert 64-bit 4CPU (+15) 42.5 − 56.5 (+27−41=31) 42.9%

but running single core, bright's results are far worse:

Thinker 5.4c Inert 64-bit (+126) 11.5 − 15.5 (+8−12= 7) 42.6%
Thinker 5.4a Inert 64-bit (+102) 16 − 34 (+8−26=16) 32.0%
Thinker 5.3b Inert 64-bit (+ 86) 13.5 − 36.5 (+7−30=13) 27.0%

Of course, I'd welcome suggestions on how to improve bright's performance against thinker, any ideas?

Nice games by Bright 0.4a, but Thinker 5.4c makes it look like an amateur.


Engine Score SP Br S-B
1: SP-x86-Inert---Thinker 15.0/18 ·················· 111111111010011111 45.00
2: Bright-0.4a 3.0/18 000000000101100000 ·················· 45.00


E-Mail Address:

As long as Thinker keeps hiding its PV and all other useful infos, I will always prefer Bright over Thinker.
In fact, my BigLion is a much better choice for 99% of chess fans than Thinker.

Matthias.

[Allard Siemelink]

Naturally I'd like to fix all FRC bugs, could you perhaps e-mail me any problem games and/or logs you might have?

I've played 200 FRC games now under ChessGUI and it is running perfectly. Not sure what the earlier problem may have been last year. Hopefully it will continue to be fine

glad to hear it
Looking forward to bright's rating on CCRL/FRC...

[Dr.Wael Deeb]

Although I do not believe that we can conclude too much from a mere 18 games, I must admit that I too noticed that Bright seems to have trouble playing Thinker.

according to ccrl 40/4, the 4 cpu versions perform relatively well:

Thinker 5.4c Inert 64-bit 4CPU (+45) 44 − 56 (+26−38=36) 44.0%
Thinker 5.4a Inert 64-bit 4CPU (+20) 51.5 − 48.5 (+34−31=35) 51.5%
Thinker 5.3b Inert 64-bit 4CPU (+15) 42.5 − 56.5 (+27−41=31) 42.9%

but running single core, bright's results are far worse:

Thinker 5.4c Inert 64-bit (+126) 11.5 − 15.5 (+8−12= 7) 42.6%
Thinker 5.4a Inert 64-bit (+102) 16 − 34 (+8−26=16) 32.0%
Thinker 5.3b Inert 64-bit (+ 86) 13.5 − 36.5 (+7−30=13) 27.0%

Of course, I'd welcome suggestions on how to improve bright's performance against thinker, any ideas?

Nice games by Bright 0.4a, but Thinker 5.4c makes it look like an amateur.


Engine Score SP Br S-B
1: SP-x86-Inert---Thinker 15.0/18 ·················· 111111111010011111 45.00
2: Bright-0.4a 3.0/18 000000000101100000 ·················· 45.00


E-Mail Address:

As long as Thinker keeps hiding its PV and all other useful infos, I will always prefer Bright over Thinker.
In fact, my BigLion is a much better choice for 99% of chess fans than Thinker.

Matthias.

A deep look into the games will be a good start for pointing the black spots....
I am willing to help as always....
Dr.D

[Matthias Gemuh]

A deep look into the games will be a good start for pointing the black spots....
I am willing to help as always....
Dr.D

Hi Dr.,
today's strong engines all use some witchcraft called magic numbers.
Though I have not been initiated into it, I hope to improve BigLion soon.
I don't regret however that I spend my time on ChessGUI
Best,
Matthias.

[Matthias Gemuh]

Naturally I'd like to fix all FRC bugs, could you perhaps e-mail me any problem games and/or logs you might have?

I've played 200 FRC games now under ChessGUI and it is running perfectly. Not sure what the earlier problem may have been last year. Hopefully it will continue to be fine

Hi Allard,
can you check whether BrightFRC 0.4a rejects b1c1 in this position ?
b1a1 is of course stronger.

[d]rk1b2rq/pppn3p/3pp3/4p1p1/4P1n1/2PPBNN1/PP3PPP/RK4RQ w GAga -