call for participation in small UCI tournament

[abik]

I in a few days I plan to run a small tournament with UCI engines to warm up the bits of my new machine under the following conditions:

Intel Core 2 Quad Q6700 2.66GHz,
8MB L2 Cache, FSB 1066MHz, 6GB RAM
64-bit Vista Service Pack 1

Nalimove endgame tablebases (complete 3,4,5-piece, some 6-piece)
Available opening book: Chessbase Powerbook 2002 (tournament setting), or OWN
Time: 10'/40 + 10'/40 + 10'/40
Maximum 2 threads per engine, ponder ON (permanent brain)
Hash table: up to 2048MB (send me "best" setting for engine)
Nalimov cache: 32MB

I have some candidates already and will also pick some random engines rated close to BikJump, but if you are interested in participating, please drop me an email (see www.aartbik.com for contact information or use the messaging system). Especially 64-bit UCIs are welcome (experimental versions too). In addition, Fritz 11 and the Turing engine will participate to cover the top and bottom region

Aart

[IWB]

Hello

Nalimove endgame tablebases (complete 3,4,5-piece, some 6-piece)

ponder ON (permanent brain)

Aart

Please keep in mind that you may run into problems with ponder on AND Tbs as both engnes might want to access different tablebases at the same time. Just a miner glitch, but to be considered.

For my own ponder on list I decided to use just 4 pc Tbs and 64MB of cache for them. This means that there are just a few accesses and then the whole 4pc Tbs is in the memory for each engine!

Bye
Ingo

[jdart]

I'm not entirely sure what you are looking for but a few suggestions:

Arasan has a 64-bit version and is UCI compatible (http://www.arasanchess.org). Glaurung is a very strong UCI engine. Ufim (http://wbec-ridderkerk.nl/html/details1/Ufim.html) is a medium-strength engine that is stable. Others are AnMon, SOS.

--Jon

[abik]

Please keep in mind that you may run into problems with ponder on AND Tbs as both engnes might want to access different tablebases at the same time. Just a miner glitch, but to be considered.

Thanks for the heads-up Ingo! This glitch actually seems to suggest a minor UCI extension for same-machine tournaments. Besides the regular "go ponder" when running on a machine exclusively, perhaps also a "go ponder_notb" would be useful, which starts pondering with tablebases access disabled until the opponent moves. Or perhaps something like "go ponder_nice" which reduces machine resource access somewhat to favor the opponent's computations, e.g. disabling tablebases access and multi-threading until it is one's own move again.

[abik]

I'm not entirely sure what you are looking for but a few suggestions.

Actually, I am not entirely sure what I am looking for myself But seriously, thanks for the suggestions. I actually wanted to start a fun tournament out of the ordinary where authors can test one or two experimental versions using multi-threading and 64-bit. The first 64-bit BikJump was cross-compiled and others very kindly tested this for me. Now that I have a 64-bit machine at home, I just wanted to return a small favor to the chess community outside the scope of the -much appreciated- serious ongoing tournaments.

[Zach Wegner]

Hello Aart,

I'd like to participate. My engine ZCT used BikJump 1.6.1 as a sparring partner and was slightly worse, until a few bugs were fixed. It should be a good competition with your newest versions. ZCT is hybrid Xboard/UCI, but the UCI support is rather rudimentary: no options, and the time control support only looks at the time remaining. Latest development versions support multiple processors, but aren't Windows compatible at the moment. I could get one to you soon though if you like (just the source--I don't have a compiler for Windows). Alternatively I could just send you the latest sources and have ZCT run on a single processor.

P.S. I'm quite jealous of your job. I'm a student in the Summer of Code this year though, so maybe it's not too far off...

[abik]

Absolutely, ZCT is very welcome to participate! Either point me to a binary you like me to test, or send me sources and I will try to get a working binary for you (I have the Intel 10.0 32-bit and 64-bit compiler or MS Visual Studio C++ .NET 2003).

I love my job. It is a really cool place to work!

So far I have the following engines (either by request or selected at random by me):

1 Mediocre v0.332
2 Roce 0.0380
3 Predateur v0.1.5
4 Philou 1.0.3
5 Monarch 1.7
6 Fritz 11
7 Arasan 10.4 (SMP)
8 BikJump v1.8 (64-bit)
9 Gibbon 2.42c
10 Turing

A test tournament is currently ongoing, posted on
http://www.aartbik.com/MISC/TOUR/

[Zach Wegner]

Absolutely, ZCT is very welcome to participate! Either point me to a binary you like me to test, or send me sources and I will try to get a working binary for you (I have the Intel 10.0 32-bit and 64-bit compiler or MS Visual Studio C++ .NET 2003).

I love my job. It is a really cool place to work!

So far I have the following engines (either by request or selected at random by me):

1 Mediocre v0.332
2 Roce 0.0380
3 Predateur v0.1.5
4 Philou 1.0.3
5 Monarch 1.7
6 Fritz 11
7 Arasan 10.4 (SMP)
8 BikJump v1.8 (64-bit)
9 Gibbon 2.42c
10 Turing

A test tournament is currently ongoing, posted on
http://www.aartbik.com/MISC/TOUR/

Hey Aart,

Thanks for the very quick response! It's looking like a fun tournament--some engines ZCT will get slaughtered by, and others that ZCT can slaughter. You can start with the binaries compiled by Jim Ablett: http://homepages.tesco.net/henry.ablett/jims.html (scroll a little ways down, or search for "zct"). I'll try to get a Windows-compatible version up on Sourceforge in the next couple of days for you that will use SMP and might be a bit stronger anyways.

[abik]

Findings so far:
(1) Arasan 10.4 (SMP, 64-bit version) seems strong but crashed several times and lost on time, and then hangs the tournament, so I will take it out.
(2) Gibbon2.42c was not making a move right before a mate in one against Turing, and lost on time.

Still ongoing:
http://www.aartbik.com/MISC/TOUR/

[abik]

Test tournament http://www.aartbik.com/MISC/TOUR/ with timing 1'+1'' done. Results below. To be honest, BikJump did better than I expected. I can post the PGN if there is interest.

Code: Select all

                           1 2 3 4 5 6 7 8 9 0 
1   Fritz 11               * 1 1 1 1 0 1 1 1 1   8.0/9
2   BikJump v1.8 (64-bit)  0 * ½ 0 1 ½ 1 1 1 1   6.0/9  21.25
3   Mediocre v0.332        0 ½ * 0 ½ 1 1 1 1 1   6.0/9  20.75
4   Gibbon 2.42c           0 1 1 * 1 1 0 1 0 0   5.0/9
5   Monarch 1.7            0 0 ½ 0 * 0 1 1 1 1   4.5/9
6   Arasan 10.4 (SMP)      1 ½ 0 0 1 * 0 0 1 0   3.5/9  18.50
7   Predateur v0.1.5       0 0 0 1 0 1 * ½ 0 1   3.5/9  12.25
8   Philou 1.0.3           0 0 0 0 0 1 ½ * 1 1   3.5/9  10.25
9   Roce 0.0380            0 0 0 1 0 0 1 0 * 1   3.0/9
10  Turing                 0 0 0 1 0 1 0 0 0 *   2.0/9