New Version of Maverick

[Steve Maughan]

I've added some basic chess knowledge and search heuristics to Maverick (my UCI engine). This has added about 150+ ELO. Maverick now beats Monarch chess in a head to head.

You can find out more and download it here:

http://www.chessprogramming.net/compute ... -maverick/

Any feedback appreciated!

Steve

[Graham Banks]

I've added some basic chess knowledge and search heuristics to Maverick (my UCI engine). This has added about 150+ ELO. Maverick now beats Monarch chess in a head to head.

You can find out more and download it here:

http://www.chessprogramming.net/compute ... -maverick/

Any feedback appreciated!

Steve

Looking forward to trying it out!

[Martin Thoresen]

Great progress Steve, get SMP working with 8-16 cores for nTCEC Season 3 (will start in 3-4 months) and I'll guarantee you a spot!

[Steve Maughan]

Thanks Martin - when Maverick hits 3000 ELO I'll consider SMP

Steve

[Steve Maughan]

The latest version (0.1) has just scored 70% against Gibbon (CCRL rating is 2261) in my turbo testing of 300 games at 60 moves in 20 secs.

Maverick may be a little stronger than I initially expected.

I'd be interested if anyone has any test results.

Steve

[Dr.Wael Deeb]

Thanks Martin - when Maverick hits 3000 ELO I'll consider SMP

Steve

Thanks Steve & good luck
Dr.D

[Steve Maughan]

It turns out version 0.1, released earlier today, has a time management bug when playing game in "x" minutes plus increment of "y" seconds per move. In some cases it could lose on time.

So I've fixed it and released it as version 0.11. Apart from the bug there isn't any strength altering changes.

You can download from here:

http://www.chessprogramming.net/downloads/

Best,

Steve

[wolfman]

The result of my FTS shows a big jump forward.

Rank Engine Version C TSR R 40/4 %MM AAD ATMM ART
87 Maverick 0.11 64 i 2219 11.83 8.41 2.89 916
93 Maverick 0.05 64 i 2061 9.73 7.18 3.24 934

Eran

[Steve Maughan]

Thanks Eran - that's in line with my tests, although I believe the CCRL scale is about 50 ELO lower that yours. I guess hash table and null move do make a significant impact!

Best,

Steve

[emadsen]

It turns out version 0.1, released earlier today, has a time management bug when playing game in "x" minutes plus increment of "y" seconds per move. In some cases it could lose on time.

That explains the discrepancy between others' test results and mine. I usually test using a time increment. I've been so busy the last week I never bothered to examine the games. I did not realize many of the losses were due to time forfeit.

I'd be interested if anyone has any test results.

I had Maverick 0.11 play against two versions of MadChess- the latest release and a development version. I got these results:

Code: Select all

2 min + 1 sec / move
128 MB hash
CuteChess CLI

Rank Name                Elo    +    - games score oppo. draws
   1 MadChess 1.3.04    2114   20   19   500   59%  2049   12%
   2 MadChess 1.2       2050   19   19   500   50%  2049   17%
   3 Maverick 0.11 x64  2049   14   14  1000   45%  2082   15%

15 games forfeited on time- all by Maverick.

Very close! Nice progress.