About this Blog

This blog is describing the development of Alaric.

New Version

blogPosted by Peter Fendrich Tuesday, January 20 2009 17:48:53
Well, I promised before new year but that was a little too early....
I actually have a new version. It is SMP and with some other changes.

It still is a little shaky though, so I will test it further a little while and then call for beta testers.
It should work with 4 threads but I am not sure what will happen with more than that. I don't have more cpu's in my computers!

  • Comments(7)//

Book problem

blogPosted by Peter Fendrich Wednesday, November 26 2008 23:39:40
I got an email from Guenther about a problem with the book. Alaric played a move that is blocked from the book. I cannot repeat the same error and cannot find out in the code how it could happen. One suggestion was that the learning file overrides the book but from what I understand it is not possible. The only situation that a blocked move will be played by Alaric is when all the legal moves are blocked by the book. Then Alaric must play something...

If you find similar problems with the book please let me know. Winboard logs or similar is a big help.

The book version at Alaric's download page (for Alaric 707) is kind of old so I will update it with the latest book version. Thank you Guenther!

There is a problem with the learning file when I play Alaric vs Alaric. I know the reason but have not fixed it and it can't have anything to do with the book problem here.

BTW. I am working with the SMP version that should be out before new year smiley

  • Comments(1)//

Alaric's results

blogPosted by Peter Fendrich Monday, June 23 2008 23:52:02
I must say I am quite happy with the result in the World Rapid Computer Chess Championchip. Number 4 is even better than the year before and surrounded by multi-core programs make me feel even better smiley
Of course it is time to make Alaric multi-core as well but I wish I wasn't so lazy...

The new Rybka version seems to be better than ever before and just lost 1/2 points in 14 rounds - Wow!
Rybka.............. 13.5
Sjeng............... 10.0
Hiarcs8x........... 10.0
Alaric................ 9.5

For the whole list click here and remember that Alaric played with the handle "TerraPi"

I will give some comments to Alaric's games later if get some time.

  • Comments(2)//

World Computer Rapid Chess Championships

blogPosted by Peter Fendrich Monday, April 07 2008 12:51:16

I look forward to the World Computer Rapid Chess Championships in June 21/22.

Hopefully I have a new version until then.
Did some minor changes to 707 sofar but I want to reduce the tree even more agressive.
Maybe it is possible, maybe not...

  • Comments(1)//

I am so rusty...

blogPosted by Peter Fendrich Tuesday, February 12 2008 23:09:19
I finally opened up my program and started to look around to remember what I once did and why. Hmmmm, did I really write that code??? Why can't I understand what this or that does?
Anyway, chess programming is starting to feel fun again and hopefully I will know what I am doing after a while smiley

I was in the middle of the endgame rules/tables ideas when I stopped everything. I will take that up later again. Right now I have some some fresh ideas in the search to try out first.
It is very tempting to have a multi-cpu version to.

  • Comments(1)//

Did the Hiarcs team cheat?

blogPosted by Peter Fendrich Friday, July 27 2007 16:38:25
The Hiarcs team changed the contempt factor during the game versus Alaric.
The contempt factor is a way to tell the engine when it is time to accept a draw. By setting the contempt factor to 1 pawn they told Hiarcs to accept a draw only if the evaluation said something worse than 1 pawn. Look at draw 149 in the game below. Hiarcs sacs a pawn in order to avoid the 50-move rule because of the high contempt factor. Hiarcs rather lose a pawn than accepting a draw.

Now, to change parameters or to intervene in any form during the game is not allowed in computer tournaments except from a few obvious cases. For instance when the engine crash you must start it up again. Normally you are not allowed to help the engine in any way.

So obviously the Hiarcs team broke that rule and they did also very openly admit that. This started a hot debate at CCC. It all starts here and the following thread.

Let us try to have balanced view here... smiley
ICC tournaments are always done in a friendly atmosphere. There are rules and there are rules. Many rules are overruled when someone accidently do some error. For instance in one game Crafty didn't Kibitz. They whispered instead, by mistake of course. The rule says kibitzing but this is typical a situation when no one tries to forfeit Crafty. Some games are replayed cause of some mistake from one of the participants. In this tournament Erdo registered at the same day. A bit late... Operators often didn't show up in time to the rounds and the TD started their engines for them. There are probably other examples as well. The TD did a great job! and no one really cared about all this. On the contrary if I illegaly tried to use Rybka and call it Alaric, there would be no friendly atmosphere anymore! Absolutely forbidden of course. There are rules and there are rules.

The friendly way gives a fun and relaxed tourney. The problem with friendlyness is that we have a gray zone of breaking rules. For instance Rybka didn't Kibitz at all in any game. The rules are very clear about this. Of course Vas could implement Kibitzing in his engine in less than 30 minutes but he didn't. Now Vas told us that he tried to use Chess Assistant. It kibitzes automatically for the engine, but he couldn't make it work and had to use Arena. In Arena the engine must kibitz by itself. Most people accepted this but there were discussions. What would happen if it was Alaric and not Rybka who did this? Would they let Alaric play? Well, I don't know. Somewhow I have a feeling that the top engines have more bargaining room when it comes to these matters.
Rybka won the tourney despite breaking this rule.

So how should we judge the Hiarcs contempt? First of all they didn't gain anything from this, they lost instead of a draw, but this is not the issue here. Secondly, if someone do change the contempt during the game, we can never never see that or prove that in any way. The same goes for many engine parameters and even things like forcing the engine to think more or to move immediately. Forbidden but not possible to discover. Is breaking a rule less serious when it can not be proven? Of course not! We would actually need a special GUI that blocks or reports any operator action during the game and that is secured in some way from hacking. We dont' have that but maybe something to think about in the future!

IMO is changing the contempt factor during the game a more serious act than the Rybka mistake, but....
The friendly atmosphere how about that?
The Hiarcs team admitted by themself what they did. They could be quiet about it.

Let us keep the friendly atmosphere even here. Let it be...
and remember that I would say the same even if Hiarcs won the game.

  • Comments(0)//

Alaric @ WCRCC tournament

blogPosted by Peter Fendrich Tuesday, July 24 2007 01:39:46

There was a World Computer Rapid Chess Championships played at ICC 21-22 July

Alaric ended up at the 6th place out of 41 engines!

All the results can be found here (look for TerraPi).

This is a really encouraging result for Alaric smiley taking to account that all engines above Alaric and some of them below are all multi-CPU engines. Alaric was playing with only one processor.
The book from G√ľnther Simon worked fine. There were some black positions that Alaric didn't know how to handle and the last game vs Hiarcs was a win much thanks to this book.
In general the tourney was excellent handled despite som small incidents. For instance engines not online when the round was to start and that Rybka didn't kibitz (display) the thinking lines.
Kibitzing was one of the rules for this tourney.
Rybka was the winner this time and in the last round Hiarcs missed the playoff by losing to... right Alaric!
A pity of course, a playoff between Rybka and Hiarcs would be a great event! But I can't say that I am extemely sorry smiley

Let's take look at the game against Quark in round 12. Alaric i black:

1. d4 Nf6 2. c4 e6 3. Nc3 d5 4. Nf3 Be7 5. Bg5 h6 6. Bh4 Nbd7 7. e3 Ne4 8. Bxe7 Qxe7 9. Rc1 Qb4 10. Qc2 Ndf6 11. Bd3 Nxc3 12. a3 Qa4 13. Qxc3 dxc4 14. Bxc4 O-O 15. O-O Bd7 16. Rfe1 Rac8 17. e4 Bc6 18. Qd3 Rfd8 19. h3 Qa5 20. b4 Qb6 21. Kh1 a6 22. Rc2 Rd7 23. Rd2 Rcd8 24. Kh2 g6 25. g3 Ba4 26. Qc3 Kg7 27. Bd3

Blog Image

Alaric evaluates this position as equal while Quark says something like 1 pawn up.
Quark values mobility high and Alaric just "notice" it. A few moves later Alaric can "see" the danger with evaluation about -0.5 and at that time Quark evaluates itself to 2 pawns up.
Quark won of course.
I have to study this more. Maybe there is something broken with Alaric's mobility evaluation?

27...Bc6 28. Bc2 h5 29. a4 Re7 30. Ne5 Be8 31. f4 Qa7 32. g4 hxg4 33. hxg4 Kf8 34. Rh1 Ng8 35. g5 Qb6 36. Kg1 f6 37. a5 Qa7 38. gxf6 Nxf6 39. Rh8+ Ng8 40. Kh1 Rg7 41. Qh3 Bf7 42. Qh4 Qxd4 43. Qxd8+ Qxd8 44. Rxd8+ Be8 {Alaric resigns}

This was the only lost game to a single CPU engine.
Ohh my, I have to start with SMP coding for multiple CPU's in order to keep up...

  • Comments(0)//

Alaric 707

blogPosted by Peter Fendrich Saturday, July 14 2007 13:00:24

Released Alaric 707 today. smiley

More at my hompage

Now I will have time to do my endgame things again!

  • Comments(2)//
Next »