GOOB 2.1 - New Version

[Gabrielmonsss]

Hi everyone,

A new version of GOOB has been released and is now available for testing

This release includes improvements to search, packed bitboards for move gen, bucketed transposition tables, along with Syzygy tablebase support, MultiPV, and NNUE(still experimental, i am yet to train the network).

In my testing against GOOB 1.8.9, the new version achieved

+178.2 ± 53.8 Elo
LOS: 100.0% | Draw ratio: 38.0%

The engine is open source and available here:

GitHub: https://github.com/Gabrielkaos/ChessEngineGoobC

If anyone from CCRL or other testing groups is interested in testing the engine, I would appreciate having this release included in a test.

Thank you.

[Gabor Szots]

Hi everyone,

A new version of GOOB has been released and is now available for testing

This release includes improvements to search, packed bitboards for move gen, bucketed transposition tables, along with Syzygy tablebase support, MultiPV, and NNUE(still experimental, i am yet to train the network).

In my testing against GOOB 1.8.9, the new version achieved

+178.2 ± 53.8 Elo
LOS: 100.0% | Draw ratio: 38.0%

The engine is open source and available here:

GitHub: https://github.com/Gabrielkaos/ChessEngineGoobC

If anyone from CCRL or other testing groups is interested in testing the engine, I would appreciate having this release included in a test.

Thank you.

Hi,

I am totally confused.

1. Where is the engine from that we have tested under the name GOOB 2.0.0? It is not in 'Releases'.
2. If you haven't trained the network, where does it come from?

[Frank Quisinsky]

Hi Gabor,

Goob 1.89 was the lastest version by the programmer on Github with executables.
The programmer created SMP support and offer the sources for v2.00 only.

I asked Jim for the compiles after.
Jim created the compiles for v2.00.

So I am thinking, if CCRL made a test ... it must be the version 2.00 from Jim Ablett.

That is really bad that in CCRL no one knows whether a JA version was used.
One of the countless things I don't like about CCRL.
The list is long ...

Best
Frank

[Frank Quisinsky]

However, the list of things I really like about CCRL is at least just as long.

[Gabor Szots]

Hi Gabor,

Goob 1.89 was the lastest version by the programmer on Github with executables.
The programmer created SMP support and offer the sources for v2.00 only.

I asked Jim for the compiles after.
Jim created the compiles for v2.00.

So I am thinking, if CCRL made a test ... it must be the version 2.00 from Jim Ablett.

That is really bad that in CCRL no one knows whether a JA version was used.
One of the countless things I don't like about CCRL.
The list is long ...

Best
Frank

Hi Frank,

I very well know it was the JA compile we tested. Now I regret having tested it, in the future I will test only executables released by the authors themselves.
I really think it a bad idea to ask Jim to compile non-official, unreleased versions.

Best,
Gabor

[Frank Quisinsky]

Hi Gabor,

I know your opinion on this.
The source code for 2.00 was available without a “dev” or “PRE” status.
So, in my opinion, it was perfectly fine to ask Jim to compile it.
Jim’s opinion on this matter is also very clear.
Furthermore interesting to see in case of GOOB ... are Jim's compile faster or not.

I have a different opinion!
If a programmer offers "executables" for a “dev” or “PRE” or "Beta" or "Alpha" or "Cesar" version … I think they should have it tested, if possible by one of the testers. Why not!

I’ve changed my mind about this situation:
The programmer made the source code available on their website / GitHub without having he created executables for a “dev” or “PRE” version themselves. That shouldn’t be tested, this is where I’ve changed my mind after about a year of discussion on TalkChess.

Best
Frank

[Gabrielmonsss]

Hi Gabor,

Goob 1.89 was the lastest version by the programmer on Github with executables.
The programmer created SMP support and offer the sources for v2.00 only.

I asked Jim for the compiles after.
Jim created the compiles for v2.00.

So I am thinking, if CCRL made a test ... it must be the version 2.00 from Jim Ablett.

That is really bad that in CCRL no one knows whether a JA version was used.
One of the countless things I don't like about CCRL.
The list is long ...

Best
Frank

Hi Frank,

I very well know it was the JA compile we tested. Now I regret having tested it, in the future I will test only executables released by the authors themselves.
I really think it a bad idea to ask Jim to compile non-official, unreleased versions.

Best,
Gabor

Hi sir Gabor,

I was also confused how GOOB 2.0.0 was in the ccrl index when i didnt make the release yet, 2.0.0 was for the changes i was making as i was improving 1.8.9 -> 2.1. I am very sorry for this, I hope I didn't cause too much problem for CCRL.

Best,
Gabriel

[Gabrielmonsss]

Hi everyone,

A new version of GOOB has been released and is now available for testing

This release includes improvements to search, packed bitboards for move gen, bucketed transposition tables, along with Syzygy tablebase support, MultiPV, and NNUE(still experimental, i am yet to train the network).

In my testing against GOOB 1.8.9, the new version achieved

+178.2 ± 53.8 Elo
LOS: 100.0% | Draw ratio: 38.0%

The engine is open source and available here:

GitHub: https://github.com/Gabrielkaos/ChessEngineGoobC

If anyone from CCRL or other testing groups is interested in testing the engine, I would appreciate having this release included in a test.

Thank you.

Hi,

I am totally confused.

1. Where is the engine from that we have tested under the name GOOB 2.0.0? It is not in 'Releases'.
2. If you haven't trained the network, where does it come from?

Hi sir Gabor,

As for the NNUE I am sorry if I included it in the options even though its still experimental, I am currently still coding the trainer. If GOOB 2.1 is to be tested, the default settings for GOOB 2.1 has NNUE turned off. Do not yet test NNUE for GOOB 2.1 as there is no network. I am very sorry for the confusion.

Respectfully,
Gabriel

[Gabor Szots]

As for the NNUE I am sorry if I included it in the options even though its still experimental, I am currently still coding the trainer. If GOOB 2.1 is to be tested, the default settings for GOOB 2.1 has NNUE turned off. Do not yet test NNUE for GOOB 2.1 as there is no network.
Gabriel

Hi Gabriel,

Thanks for the clarification.

In your opinion, how much version 2.1 is better than version 2.0.0?