Announcing Daydreamer 1.0

[Aaron Becker]

Daydreamer 1.0 (UCI), by Aaron Becker
http://github.com/AaronBecker/daydreamer/tree/master

I've just released the first version of Daydreamer, the chess engine I've been working on this summer. It started as a side effort in the development of GUI chess application for macs, but it was so interesting that it took on a life of its own. The engine is open source; you can take a look at the code and download binaries on the engine's site. I've only tested it extensively on a mac, but I've prepared linux and windows builds as well which at least start up correctly. Let me know if you run into any troubles.

At this point, the engine is almost all search and no evaluation. In fact, I'm using exactly the evaluation algorithm given by Tomasz Michniewski for unified evaluation tournaments. As a result, it doesn't handle complex endgames (or sometimes even simple endgames) or imbalanced positions very well, and it has no regard for king safety, pawn structure, or mobility. Even with all that, it plays a decent game; certainly it's better at chess than I am. Here are the results of quick 10 game matches with a variety of engines that I've been using to test for regressions during development, just to given an idea of where Daydreamer stands:

Code: Select all

BigLion 2.23x   +8-1=1  8.5/10
BikJump 2.01    +3-6=1  3.5/10
Clarabit 1.00   +3-4=3  4.5/10
Greko 6.5       +1-5=4  3.0/10
Mediocre 0.34   +4-6=0  4.0/10
Plisk 0.0.9     +8-0=2  9.0/10
Sungorus 1.2    +2-7=1  2.5/10

Here are its results on two common test suites at 10s per move:

Code: Select all

WAC     291/300
ECMGMP   93/173

With this release I'd like to publicly thank the authors of the open source programs that provided a lot of insights and ideas into how chess engines should be written and organized. Fruit and especially Viper and Glaurung have been particularly big influences. I also appreciate all the fine writing on chess programming that's available on the web. Bruce Moreland, Matt Gingell, Jonatan Pettersson, Tord Romstad, and the contributors to the Chess Programming Wiki have all influenced Daydreamer with their articles. Thanks everyone, it's been fun so far.

[Leo_D]

Hi Aaron,

Thanks for making Daydreamer available!
One little question, what country it comes from?

Best wishes,
Leo
http://wbec-ridderkerk.nl/html/newspage.htm

[Aaron Becker]

Hi Leo, Daydreamer and I are from the USA. Thanks for adding Daydreamer to your page, I appreciate the publicity.

[swami]

Any direct link to the windows compile?

[Aaron Becker]

The windows compile can be found under the downloads tab in github, or directly at http://cloud.github.com/downloads/Aaron ... reamer.exe.

I'm interested to hear reports from windows users; I've never cross-compiled for windows before, and I've only done a small amount of testing in a virtual machine.

[swami]

The windows compile can be found under the downloads tab in github, or directly at http://cloud.github.com/downloads/Aaron ... reamer.exe.

I'm interested to hear reports from windows users; I've never cross-compiled for windows before, and I've only done a small amount of testing in a virtual machine.

Thank you, It's working just fine. If you'd like to have the best compile for your engine, Jim Ablett can always make one for you. I feel this compile from yours is as good as what Jim would usually have for many engines.

It's analysis lines looks unique.

[Aaron Becker]

Thank you, It's working just fine. If you'd like to have the best compile for your engine, Jim Ablett can always make one for you. I feel this compile from yours is as good as what Jim would usually have for many engines.

It's analysis lines looks unique.

Glad to hear it's working. There's a lot of weird stuff in its output that's probably only interesting to me, left over from some marathon debugging sessions. I thought about removing them all before releasing it, but I sort of like it the way it is. It follows the uci spec, so I'm hoping it won't cause any problems, but I'm a little paranoid about how other GUIs will treat it.

Unless you meant the pvs it was coming up with look unique, in which case no comment.

[swami]

Day Dreamer 1.0

Strategic Test Suite Conditions:

Arena GUI
10 seconds per position
500 positions
Core2Quad 32 bits, Q6600, 2 GB RAM, 2.4GHZ
Engine uses 129 Mb Hash.

Subject-wise Scores:

STS (v1.0) - Undermining:
46/100, Grade: D

STS (v2.0) - Open Files and Diagonals:
34/100, Grade: ?

STS (v3.0) - Knight Outposts/Centralization/Repositioning:
41/100, Grade: E

STS (v4.0) - Square Vacancy:
47/100, Grade: D

STS (v5.0) - Bishop vs Knight:
57/100, Grade: C+

Overall Performance:

Total Score: 225/500

Overall Average: 45.0 %

Grade: D

[Volker Pittlik]

... Let me know if you run into any troubles.

I get a segmentation fault here after compiling it without errors.

volker@vpittlik:~/schach/daydreamer/1.0$ gcc --version
gcc (GCC) 4.2.4 (Ubuntu 4.2.4-1ubuntu4)

volker@vpittlik:~/schach/daydreamer/1.0$ uname -a
Linux vpittlik 2.6.24-24-generic #1 SMP Fri Jul 24 22:15:50 UTC 2009 x86_64 GNU/Linux


vp

[Aaron Becker]

I get a segmentation fault here after compiling it without errors.

volker@vpittlik:~/schach/daydreamer/1.0$ gcc --version
gcc (GCC) 4.2.4 (Ubuntu 4.2.4-1ubuntu4)

volker@vpittlik:~/schach/daydreamer/1.0$ uname -a
Linux vpittlik 2.6.24-24-generic #1 SMP Fri Jul 24 22:15:50 UTC 2009 x86_64 GNU/Linux
vp

Hmm, I'm not sure what the problem might be. I used gcc and Ubuntu for my linux testing as well, although not exactly the same versions. I think I need a little more information to find the problem.

Is this a 64-bit executable? I only tested 32-bit versions so far, so there may be problems lurking in a 64-bit build. Does it segfault as soon as you start it up, or only after it starts to search? Does it print any output at all, other than "segmentation fault"?

You can build a debugging version by running "make debug". Trying that version might shed some light on the problem. Best of all would be to run the debug version inside gdb--then when it crashes it will give a full stack trace, which should be enough to pinpoint the problem.

Sorry for the trouble, and thanks for any help you can give me in tracking the problem down.