There is a new Gull Development version

[acase]

For those who didn't know, there is a new Gull Development version as of Feb 15 2016.


For some reason though I think I might be a little bit behind the times, as everyone here seems to spot new stuff before I do.


Get the version you need here :

http://abrok.eu/gull/

[Ajedrecista]

Hello Andrew:

I do not know the amount of code that has been changed. However, there is a 1000-game match available with its corresponding PGN at ImmortalChess Forum:

Gull 3 Syzygy vs Gull 16021418 x64 BMI2 - 1000 Games

Code: Select all

Games Completed = 1000 of 1000 (Avg game length = 19.079 sec)
Settings = RR/32MB/Book/6000ms+60ms/M 1000cp for 12 moves, D 150 moves/PGN:H:\LittleBlitzer-2.75\2moves_v1.pgn(32000)
Time = 4951 sec elapsed, 0 sec remaining
1. Gull 3 TB Syzygy 486.5/1000 217-244-539 (L: m=136 t=0 i=0 a=108) (D: r=370 i=96 f=33 s=0 a=40) (tpm=124.0 d=12.69 nps=2820618)
2. Gull 16021418 x64 BMI2 513.5/1000 244-217-539 (L: m=117 t=0 i=0 a=100) (D: r=370 i=96 f=33 s=0 a=40) (tpm=123.7 d=12.83 nps=2987219)

I obtain a measured Elo gain of 9.4 ± 14.6 Elo = [-5.2, 24] Elo with 95% confidence and a LOS value of 89.6% more less. It is not very significant for me. Anyway, thanks for the info. I wish good luck to Vadim in the development of Gull.

Regards from Spain.

Ajedrecista.

[duncan]

For those who didn't know, there is a new Gull Development version as of Feb 15 2016.


For some reason though I think I might be a little bit behind the times, as everyone here seems to spot new stuff before I do.


Get the version you need here :

http://abrok.eu/gull/

[d]3B4/8/8/8/Bk6/8/8/7K w

stockfish finds a mate (not the fastest) in 10 seconds for the above position on my computer (longest bishop mate)

gull after 5 minutes still has not found it. any ideas why?

[acase]

For those who didn't know, there is a new Gull Development version as of Feb 15 2016.


For some reason though I think I might be a little bit behind the times, as everyone here seems to spot new stuff before I do.


Get the version you need here :

http://abrok.eu/gull/

[d]3B4/8/8/8/Bk6/8/8/7K w

stockfish finds a mate (not the fastest) in 10 seconds for the above position on my computer (longest bishop mate)

gull after 5 minutes still has not found it. any ideas why?

Hello Duncan,

I have no idea why, I could mate from that position myself, however it appears some programs don't have that built into them, and Gull doesn't support any tablebases that I know of.


Thanks for pointing that out by the way, I find it interesting.

[Leto]

There's a Gull that supports syzygy out there, was released recently.

[Gusev]

Great news! Will there be also a GullTest, an analog of FishTest?

For those who didn't know, there is a new Gull Development version as of Feb 15 2016.


For some reason though I think I might be a little bit behind the times, as everyone here seems to spot new stuff before I do.


Get the version you need here :

http://abrok.eu/gull/

[acase]

Hello Andrew:

I do not know the amount of code that has been changed. However, there is a 1000-game match available with its corresponding PGN at ImmortalChess Forum:

Gull 3 Syzygy vs Gull 16021418 x64 BMI2 - 1000 Games

Code: Select all

Games Completed = 1000 of 1000 (Avg game length = 19.079 sec)
Settings = RR/32MB/Book/6000ms+60ms/M 1000cp for 12 moves, D 150 moves/PGN:H:\LittleBlitzer-2.75\2moves_v1.pgn(32000)
Time = 4951 sec elapsed, 0 sec remaining
1. Gull 3 TB Syzygy 486.5/1000 217-244-539 (L: m=136 t=0 i=0 a=108) (D: r=370 i=96 f=33 s=0 a=40) (tpm=124.0 d=12.69 nps=2820618)
2. Gull 16021418 x64 BMI2 513.5/1000 244-217-539 (L: m=117 t=0 i=0 a=100) (D: r=370 i=96 f=33 s=0 a=40) (tpm=123.7 d=12.83 nps=2987219)

I obtain a measured Elo gain of 9.4 ± 14.6 Elo = [-5.2, 24] Elo with 95% confidence and a LOS value of 89.6% more less. It is not very significant for me. Anyway, thanks for the info. I wish good luck to Vadim in the development of Gull.

Regards from Spain.

Ajedrecista.

Thanks for the info Jesus, one of the differences I've noticed with this new development version is that it now supports more than 16 cores (unlike Gull3).


Regards from Columbus, Ohio USA

[velmarin]

Thanks for the info Jesus, one of the differences I've noticed with this new development version is that it now supports more than 16 cores (unlike Gull3).
Regards from Columbus, Ohio USA

You only have to change the variable or maximum _cpus definition,
Another thing is to have cpus on the computer to test it.???
For something it starts...
But Gull not is stockfish, it seems more complicated, it seems to me humbly.

[Gusev]

This feature was likely inherited from Gull 3.0.1. It is necessary for MinGW compatibility, because the function that was used in Gull 3 to block the increase of the number of threads beyond the number of cores is not supported in MinGW. But neither is it needed, so this is okay.

Hello Andrew:

I do not know the amount of code that has been changed. However, there is a 1000-game match available with its corresponding PGN at ImmortalChess Forum:

Gull 3 Syzygy vs Gull 16021418 x64 BMI2 - 1000 Games

Code: Select all

Games Completed = 1000 of 1000 (Avg game length = 19.079 sec)
Settings = RR/32MB/Book/6000ms+60ms/M 1000cp for 12 moves, D 150 moves/PGN:H:\LittleBlitzer-2.75\2moves_v1.pgn(32000)
Time = 4951 sec elapsed, 0 sec remaining
1. Gull 3 TB Syzygy 486.5/1000 217-244-539 (L: m=136 t=0 i=0 a=108) (D: r=370 i=96 f=33 s=0 a=40) (tpm=124.0 d=12.69 nps=2820618)
2. Gull 16021418 x64 BMI2 513.5/1000 244-217-539 (L: m=117 t=0 i=0 a=100) (D: r=370 i=96 f=33 s=0 a=40) (tpm=123.7 d=12.83 nps=2987219)

I obtain a measured Elo gain of 9.4 ± 14.6 Elo = [-5.2, 24] Elo with 95% confidence and a LOS value of 89.6% more less. It is not very significant for me. Anyway, thanks for the info. I wish good luck to Vadim in the development of Gull.

Regards from Spain.

Ajedrecista.

Thanks for the info Jesus, one of the differences I've noticed with this new development version is that it now supports more than 16 cores (unlike Gull3).


Regards from Columbus, Ohio USA

[Dann Corbit]

It is not as complicated as it seems.
All of the code in one module makes it a bit daunting, but it could be split.

There is a huge material imbalance table. I think that if you carefully study how it is initialized you will find that part fascinating.

Consider how strong it is, with only one person working on it.
If it should receive the stockfish gang of many treatment, it will become the world's strongest.

It has all the best concepts from Ivanhoe with all the bugs removed. All of the less useful stuff has been left out. I don't think that Ivanhoe can go very far, though Firebird/Fire has gone forward some.

If nothing else, it should be carefully studied. All the great ideas of Glaurung moved forwards into Stockfish and look what became of it.

If the same thing should happen to Gull, it would be a mighty hammer indeed.

Besides, being open source, I can modify it for my own evil purposes.