Prédateur 2.1 released, now x32, x64 and Mac!

[JuLieN]

Prédateur 2.1 released, now x32, x64 and Mac!

This new version of Prédateur wasn't meant to be public, as it is only slightly stronger than the previous version (maybe +50 Elo), while I prefer to release new versions when they are significantly stronger than older ones. But it turns out that I started to rewrite Prédateur using C++ and bitboards, so I had this last pascal version sitting in my harddrive, with some improvements and bugfixes and hence decided to release it anyway. Also, it now comes in three flavors : Win32, Win64 and MacOSX (the Mac version can't "listen" to commands while thinking, so I disabled the ponder mode or it would go into an endless search, unable to answer to a "ponderhit" command). The Mac version will work with Sigma Chess and Jose, but will make Stockfish's interface crash, for some reason I don't know.

Enjoy this new release, before the new generation of Prédateur.

Note: this 2.1 version of Prédateur is not to be confused with the "2.1" that entered Olivier Deville's latest OpenWar, which should actually be seen as "2.0.1".

What changed since last update ?

- Prédateur now prunes a tad less, but (hopefully) in a smarter way : it should diminish search instability and weird moves that sometimes happened.
- memory-overuse by hash tables fixed (thanks, Joost!)
- in a general way, diminution of the memory footprint : Prédateur now uses "only" 6 mb of non-hash memory.
- speeded up the move generator, especially the check routine
- some evaluation modifications: Prédateur should now try to target your king a bit more... be careful!
- Mac OSX version. (Ponder mode disabled).
- some other bugs fixed, here and there.

Blog release : http://predateur-chess.blogspot.com/201 ... 4-and.html
Download archive : http://julien.marcel.free.fr/public/Predateur2.1.rar

[imai]

JulieN, should this version also run in Xboard on a Mac ?
When I unrar, Predateur_2.1_mac will appear as a document in Mac, not as a Unix executable file. It would be nice if it could run in Xboard !

[Josh Pettus]

Try
"sudo chmod 755 predateur"
in a terminal

Generally need to do that when you dl unix executables from over the web.

[imai]

Yes, that works Predateur is running in Xboard. Thanks

[JuLieN]

Ouch, I totally missed this thread development, sorry. Thanks for having answered imai's question, Darklord. Yes, the executable's bit was lost with the win32 archiver I used, so one has to add it back with a "chmod +x filename". I'll archive the next releases directly from MacOSX, so there shouldn't be anymore problems of that kind.

Hopefully you're glad with Prédateur, imai!

[ZirconiumX]

Evidently the mac binary is Intel only as, well:
[quote='Terminal']
-bash: /Users/***/Desktop/Predateur2.1/Predateur_2.1_mac: cannot execute binary file
[/quote]

PM me the source and I'll see what I can do.

Matthew:out

[JuLieN]

Evidently the mac binary is Intel only as, well:
[quote="Terminal"']
-bash: /Users/***/Desktop/Predateur2.1/Predateur_2.1_mac: cannot execute binary file
[/quote]

PM me the source and I'll see what I can do.

Matthew:out[/quote]

Hello Matthew,

thanks for your "proposal" (although it looks more like an order to me... ), but:

1) My engine is closed-sources.
2) Compiling a PPC or universal binary is a trivial task that only needs to check a box in the compiler settings. So if I wanted to produce one I thiiiink I wouldn't need to rent an expert

[imai]

JuLieN,
Yes I like Predateur, it gives me the illusion that I can beat it but up to now it still is an illusion.
Looking forward to your next release.
All the best

[JuLieN]

Thx imai! The only reward a programmer gets from a free program is satisfied user! ^^

BTW, Prédateur can be set to "depth" levels of difficulty (for instance: it can be set to think at a maximal depth of 1 ply, 2 plies, etc...). You can set that in most user interfaces. For instance, with Arena:

- go to the "level" menu
- chose "adjust"
- then check the box called "Fixed search depth" and set it to "1" in the right panel.
There you go!

You can safely start at a depth of "1 ply": you'll see that Prédateur still makes for a tough challenge. (The first time I tried it at this depth it got me surprised and I lost!). When you tame it at this depth, you can proceed and increased the search depth, and so on... Also, at such shallow depths it answers nearly instantly (on my Core i5, Prédateur takes less than 200 milliseconds to reach a depth of 7-8 plies).

EDIT: Here's a nice game I just played at depth 1. As you can see, and despite I won, Prédateur still plays a good game

Code: Select all

[Event "?"]
[Site "?"]
[Date "2011.07.15"]
[Round "?"]
[White "Julien"]
[Black "Predateur_22271_11"]
[Result "1-0"]

1. e4 c6 2. Nf3 d5 3. exd5 Qxd5 {(Dxd5) +0.04/1 0} 4. d4 Nf6 5. Nc3 Qf5
{(Df5) 0.00/1 0} 6. Bd3 Qe6+ {(De6+) -0.17/1 0} 7. Be3 Bd7 {(Fd7) -0.20/1
0} 8. O-O Na6 {(Ca6) -0.25/1 0} 9. Re1 O-O-O {(0-0-0) -0.08/1 0} 10. Bxa6
bxa6 {(bxa6) +0.13/1 0} 11. Na4 Ne4 {(Ce4) +0.19/1 0} 12. Bf4 Qd5 {(Dd5)
+0.19/1 0} 13. c3 Bf5 {(Ff5) +0.12/1 0} 14. Ne5 Rg8 {(Tg8) +0.17/1 0} 15.
f3 Nd6 {(Cd6) +0.06/1 0} 16. Nc5 Nc4 {(Cc4) 0.00/1 0} 17. Qa4 Nxb2 {(Cxb2)
+0.01/1 0} 18. Qxa6+ Kc7 {(Rc7) -0.98/1 0} 19. Ned7+ e5 {(e5) -1.03/1 0}
20. Bxe5+ Bd6 {(Fd6) -1.00/1 0} 21. Qb7# 1-0

[imai]

Txs JuLieN,
Reduced strength is a good feature that should be part of every engine. There are not that many chess players who can beat the computer nowadays.
However, on my Macbook I have only 2 GUI's , Stockfish and Xboard. Configuration of Predateur in Stockfish is limited to the hash value and choice of the opening book, in Xboard you can only set the Polyglot items, so sadly it is not working for me.
I'll have to ask Tord Romstad and H.G.Muller to make this available.
All the best

[JuLieN]

Txs JuLieN,
Reduced strength is a good feature that should be part of every engine. There are not that many chess players who can beat the computer nowadays.
However, on my Macbook I have only 2 GUI's , Stockfish and Xboard. Configuration of Predateur in Stockfish is limited to the hash value and choice of the opening book, in Xboard you can only set the Polyglot items, so sadly it is not working for me.
I'll have to ask Tord Romstad and H.G.Muller to make this available.
All the best

I have a Mac too, and that's what I use to develop Prédateur. May I suggest you to use Sigma Chess? It is, IMHO, the best interface for our platform (for the moment ), and it is now FREE!

To add and configure Prédateur, here's what you do :
-> Analyze menu
-> Engine
-> Engine Manager
-> Add

When you have installed Prédateur, you can then reduce its search to a one-ply deep one by doing the following:
-> Level
-> Playing Mode & Level
-> Fixed Depth
-> And you set it to the max depth you wish. For instance "1" to start.

You can download Sigma Chess here:
http://www.sigmachess.com/_download/download.html

SC is free, but comes with several engines, some of them being lite versions (for instance Hiarcs and Deep Junior). You can install any UCI engine compiled for Mac.

[imai]

JuLieN,
Thanks for the info. I downloaded Sigma, installed Predateur and set the mode to "Novice" and level to "Easiest"
So far so good but when I started a game this happened :

Code: Select all

[Event "?"]
[Site "?"]
[Date "2011.07.16"]
[Round "?"]
[White "?"]
[Black "Predateur v2.1 x64"]
[Result "*"]
[ECO "B02"]

1. e4 Nf6 2. Nc3 2. Bd3 *



On the second move for Predateur (black) it makes a move for white !!
I tried another opening and got the same result, black makes a move for white on the second move.
Is this a bug or is get a novice an extra move on this level ?
Best regards

[JuLieN]

Wow that's weird! It doesn't happen with Arena, so I don't know what's wrong here. And the problem is that AFAIK sigma Chess doesn't provide any log of the engine's input/ouput...

[imai]

Yes, must be a bug in Sigma. So adieu to Sigma and back to Xboard. The best equalizer in Xboard seems to be the -TimeOdds function to set the available time for the engine to a fraction of the time specified for the user.
All the best

[kinderchocolate]

Thanks for the release. It's fun to play it. Can you tell us a little about your future plans? Eg: Would you implement a skill level UCI option like what Stockfish has done. Or are you thinking of pushing it to the world computer chess championship?

[JuLieN]

Thanks for the release. It's fun to play it. Can you tell us a little about your future plans? Eg: Would you implement a skill level UCI option like what Stockfish has done. Or are you thinking of pushing it to the world computer chess championship?

My pleasure! Again, it's always nice to ear people having fun with your creation! ^^

About future plans... Prédateur is my first (serious) chess engine (Actually, the first program I ever wrote was yet a chess program, in Amiga Basic, on my Amiga 500 computer, back in 1991...), so it is kind of a naive one. It's now 3-4 years old and I believe I've pushed this naive architecture to its limits. But during those four years I've learnt a LOT about chess programming, so I believe it's time to rewrite it from scratch and make it better using my new understanding.

I've ported it last month to the iPhone, still in free pascal, but using non-Apple development tools for such platforms is really not easy. So, despite I really like the Pascal language, I've started rewriting Prédateur in Objective C. An yes, this one I'd like to be able to challenge the best engines. So it is bitboards-based (my own implementation of it, anyway, that might not be very orthodox...) and I'll try to make it a state of the art program in every way. It will be tagged Prédateur 3.0 when it will be released. In the mean time, and because I can't stand bugs and unused ideas, I'll probably release a few more 2.x versions in Free Pascal

Right now, I'm focusing on the iPhone/iPad version, as I'd like to release something on the AppStore in september.

EDIT: about skill level... I find those functions a bit artificial. It gives the engines a non-natural way of blundering. i think that a better way to handicap the program is really to reduce its search depth. And, again, even at depth 1 it still produces a surprisingly nice game to play!

[H.G.Muller]

My 1979 engine for the 6502, Usurpator II, was just able to steal a half-point from it in the ChessWar promo, against the rating odds. An uncurable mate threat forced Predateur to go for the perpetual.

Code: Select all

[Event "ChessWar XVII Promotion 40m/20'"]
[Site "DEVILLE1-5D9C28"]
[Date "2011.07.22"]
[Round "1.54"]
[White "Predateur 2.1_x64"]
[Black "Usurpator II"]
[Result "1/2-1/2"]
[TimeControl "40/1200"]
[Annotator "1. -0.34"]
[Number "54"]


1. e4 {-0.34/12} e5 2. Nf3 {+0.02/11 26} f6 3. Nc3 {+0.66/11 25} Bb4 4. Bc4
{+0.79/11 25} Ne7 5. O-O {+0.78/10 24} c5 6. d3 {+1.13/9 24} Bxc3 7. bxc3
{+0.95/10 23} Nbc6 8. Be3 {+1.00/10 23} d5 9. exd5 {+0.79/10 22} Nxd5 10.
Bxc5 {+0.79/10 22} Nxc3 11. Qe1 {+0.64/10 21} Qa5 12. Be3 {+0.48/10 21} b5
13. Bb3 {+0.98/10 20} Bg4 14. Bd2 {+0.79/11 20} b4 15. a3 {+0.59/10 19}
Bxf3 16. axb4 {+0.49/10 19} Qxb4 17. gxf3 {-0.35/10 18} Nd4 18. Bxc3
{+0.00/9 18} Nxf3+ 19. Kh1 {+0.00/7 18} Qf4 20. Ba4+ {+0.00/6 17} Kf7 21.
Bb3+ {+0.00/5 17} Ke8 22. Ba4+ {+0.00/3 17} Kf7 23. Bb3+ {+0.00/3 16} Ke8
24. Ba4+ {+0.00/3 16} Kf7
{Draw by repetition} 1/2-1/2


[JuLieN]

@HG
Hehe, well done!
Is this engine publicly available ( I LOVE vintage programs! )

[H.G.Muller]

Well, it was on my website, but I could not find any link to it! I had to look for it a bit myself, using ftp, but finally I found it at http://home.hccnet.nl/h.g.muller/usurpatorIIemu.exe .

I vaguely remember it needs a command-line argument, indicating the nominal depth in quarter plies. (It has no real time control; you have to set it to the approximate time per move you want it to play through this command-line parameter, and then hope for the best.) IIRC 12 is a good value for the TC used in the ChessWar promo. (Or was it 14? Perhaps Olivier still knows.)

It is a bit annoying that it does not produce thinking output. Originally I developed and ran it on the AIM 65 microcomputer. It did print some stats about the move, including score, in the 20-char LED display there. In the 6502 emulator that is used to run it I just ignore the subroutine calls to the ROM routines that did this printout, but perhaps I should transform it into a line of thinking output, to at least get the score. (It does not keep track of the PV.)

[Olivier Deville]

I am using 14 as a parameter.

Olivier