EXchess v6.10 release

[dchoman]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

[lucasart]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

Good stuff So you CLOP-timized your piece values. That's definitely the right approach to tuning an eval. I'll have to get round to that at some point in my engine.

I downloaded and compiled it w/o any problems on GNU/Linux, using gcc 4.6.1, as follows:

Code: Select all

g++ -o ./exchess -O3 -flto EXchess.cc

I'll test it in my Open Source Blitz rating list (cf my posts in the tournament forum), and you'll get results tomorrow.

[pocopito]

Thanks for sharing your job Dan.

It compiled fine under xubuntu 11.10, 3.0.0-15-generic 64 bit gcc-4.6.1

$ g++ -o EXchess_console src/EXchess.cc -O2


Seems to work perfectly under scidVSpc

Regards

E Diaz

[dchoman]

Hi Dan,

Nice to see you working on EXchess again.

BTW, we can omit the 'beta' tag from the name, if I understand it correctly, can't we?

Yes, the released versions of EXchess always have the 'beta' tag, so it is not a unique identifier and can be omitted if you wish.

- Dan

[dchoman]

Good stuff So you CLOP-timized your piece values. That's definitely the right approach to tuning an eval. I'll have to get round to that at some point in my engine.

I downloaded and compiled it w/o any problems on GNU/Linux, using gcc 4.6.1, as follows:

Code: Select all

g++ -o ./exchess -O3 -flto EXchess.cc

I'll test it in my Open Source Blitz rating list (cf my posts in the tournament forum), and you'll get results tomorrow.

Thanks!

The CLOP optimization is an interesting tool. I am just learning how to use it effectively. My piece + mobility values were already pretty well optimized via TD learning (or so I thought), but with about 25,000 games, CLOP changed them rather significantly. The net result was about +10 elo in my rapid game testing, so the new values make a small, but measureable improvement in play. I am not sure if I would have gotten more from 100,000 games, but I may try it at some point.

- Dan

[lucasart]

Hello Daniel,

have a look here
http://www.talkchess.com/forum/viewtopic.php?t=42230

it looks like 6.10 is a lot stronger than 6.0.2 !

congrats for the new improved release

[Graham Banks]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

Hi Dan,

do you know if Jim will be making a compile available so that those interested can test it?

Graham.

[Jim Ablett]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

Hi Dan,

do you know if Jim will be making a compile available so that those interested can test it?

Graham.

My windows compiles (mingw) are crashing on first move unfortunately

Jim.

[dchoman]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

Hi Dan,

do you know if Jim will be making a compile available so that those interested can test it?

Graham.

My windows compiles (mingw) are crashing on first move unfortunately

Jim.

Last night I found a bug in reading the score.par file. This happens during initialization, and it shouldn't cause a crash as EXchess should use the default values. But I did fix it this morning and upload a new tar ball.
Playing strength should not be affected. Not sure if this will fix the mingw problem, but I am happy to work more on it if you have a description of the problem... does it start in console mode (without winboard/xboard)?

- Dan

[Jim Ablett]

Hi Everyone,

I just put the sources for EXchess v6.10 up on my website...

http://personal.denison.edu/~homand/EXchess.htm

It has a number of minor improvements over v6.02 in king safety, some piece and mobility value optimization via CLOP, and a few minor search improvements.

I also revised the 'default' score values in the source code to match the score.par values which have been learned and hand optimized over a long period of time... this way, if score.par is deleted by accident a new version with the 'best' values will be automatically created.

- Dan

Hi Dan,

do you know if Jim will be making a compile available so that those interested can test it?

Graham.

My windows compiles (mingw) are crashing on first move unfortunately

Jim.

Last night I found a bug in reading the score.par file. This happens during initialization, and it shouldn't cause a crash as EXchess should use the default values. But I did fix it this morning and upload a new tar ball.
Playing strength should not be affected. Not sure if this will fix the mingw problem, but I am happy to work more on it if you have a description of the problem... does it start in console mode (without winboard/xboard)?

- Dan

Hi Dan,

Exchess plays first move but on opponent's move reply it crashes.

Jim.