Critter 1.0 available

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

Peter Marriott
Posts: 26
Joined: Mon May 03, 2010 7:22 am

Re: Critter 1.0 available

Post by Peter Marriott »

Thank you!
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: Critter 1.0 available

Post by rvida »

Maharadja wrote:thanks a lot, i hope the style didn't change. critter is my favourite engine :D
Playing style (as perceived by humans) is more or less only a function of evaluation parameters.

Critter 0.90 had a few 'bugs' in its evaluation which contributed to his 'playing style':
- ridiculously high bonus for rook on open file, 2x more than most engines
- endgame component of pawn chain bonus was too high
- overly aggressive 'king danger' bonus

Now in Critter 1.0 these are corrected, but still I wanted to maintain an engine with an unique and entertaining play-style, therefore my primary goals with respect to eval features were:
- don't go into unsound king attacks
- be less materialistic and more positional (see Critter's eval in Najdorf poisoned pawn variation)
- try to trade down to imbalanced endgames - Critter now often gives up a rook for a minor just to be a pawn up in the endgame, especially if the pawns are connected and not opposed

But the main deciding factor whether an eval feature stays or goes were thousands of engine-engine matches.

Richard
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: Critter 1.0 available

Post by rvida »

Don wrote: Hi Richard,

How much improved do you estimate Critter to be?

Don
My estimate is between 30-40 Elo points, depending on TC and the choice of opponents...

Richard


Edit:
I posted some unofficial results from a beta tester here on talkchess.
More official tests are in progress at http://www.inwoba.de
User avatar
Romy
Posts: 72
Joined: Thu Mar 10, 2011 10:39 pm
Location: Bucharest (Romania)

Re: Critter 1.0 available

Post by Romy »

Critter wrote:My estimate is between 30-40 Elo points, depending on TC and the choice of opponents...
From my early testworks, I agree it can be 30+ ELO. Only Houdini is usually able to contain the Critter 1.0.

My congratulation.

Soon you can be attack by baboon-pack so be careful. It is an offense to write worldclass chess program.
Tom Barrister
Posts: 227
Joined: Tue Oct 05, 2010 5:29 pm

Re: Critter 1.0 available

Post by Tom Barrister »

For some reason Critter 1.0 doesn't run on SCID-vs-PC unless I disable the opening book. 0.90 doesn't have this issue; both engines and the book are in the same directory.

I'm not complaining, since I only use engines for analysis. I'm just pointing it out in case there's some use for the information.
This production is being brought to you by Rybka: "The engine made from scratch.™"
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: Critter 1.0 available

Post by rvida »

Norm Pollock wrote:Thanks Richard.

I had to put Critter on my flash drive because that is where I have the GTBs. It seems that Critter will only use GTBs if they are on the same drive.
Now I cannot avoid tech-speak and with my very bad english it may sound rather confusing :(

This depends on the GUI used. Technically, if the GUI sends a command 'setoption name GaviotaTbPath value ...' it should work on any drive. Upon engine startup Critter tries to initialize tablebases from the '\gtb' folder on the drive it resides on. This is somewhat similar to 'AutoTb' feature of Rybka. It is the GUI's responsibility to set the correct path afterwards.

Critter 0.90 tried to re-initialize the tablebase interface after receiving 'isready' command (which is mandatory according to UCI specification). If the GUI does not send this command, or sends it before setting the GaviotaTbPath option, Critter won't find the GTB files.

Now Critter 1.0 is more liberal, if any of GTB-related options were changed after engine startup, it tries to re-initialize the tablebase interface in its own thinking time after the first 'go' command from the GUI. If this does not help then the GUI must be broken...

Richard
Matt Weaver

Re: Critter 1.0 available

Post by Matt Weaver »

Absolutely wrong. An offense is only to use someone other's code verbatim in a program, world-class or not, and declare it as your own work. Fortunately, Richard does not seem to be this type of "genius" or "saint" so it would be great one day to have a #1 with a clean record and conscience. Hopefully Richard will go that far as to succeed as Number 1.
Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: Critter 1.0 available

Post by Albert Silver »

rvida wrote:
Maharadja wrote:thanks a lot, i hope the style didn't change. critter is my favourite engine :D
Playing style (as perceived by humans) is more or less only a function of evaluation parameters.

Critter 0.90 had a few 'bugs' in its evaluation which contributed to his 'playing style':
- ridiculously high bonus for rook on open file, 2x more than most engines
- endgame component of pawn chain bonus was too high
- overly aggressive 'king danger' bonus

Now in Critter 1.0 these are corrected, but still I wanted to maintain an engine with an unique and entertaining play-style, therefore my primary goals with respect to eval features were:
- don't go into unsound king attacks
- be less materialistic and more positional (see Critter's eval in Najdorf poisoned pawn variation)
- try to trade down to imbalanced endgames - Critter now often gives up a rook for a minor just to be a pawn up in the endgame, especially if the pawns are connected and not opposed

But the main deciding factor whether an eval feature stays or goes were thousands of engine-engine matches.

Richard
I was actually fond of its attacking style. Any chance you could leave it as a parameter in the UCI?
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
User avatar
Dr.Wael Deeb
Posts: 9773
Joined: Wed Mar 08, 2006 8:44 pm
Location: Amman,Jordan

Re: Critter 1.0 available

Post by Dr.Wael Deeb »

Romy wrote:
Critter wrote:My estimate is between 30-40 Elo points, depending on TC and the choice of opponents...
From my early testworks, I agree it can be 30+ ELO. Only Houdini is usually able to contain the Critter 1.0.

My congratulation.

Soon you can be attack by baboon-pack so be careful. It is an offense to write worldclass chess program.
It looks like you're obsessed with this baboon thingy....

Hmmm,could it be that a baboon had abused you somehow in your childhood :!: :?:

:roll:
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: Critter 1.0 available

Post by Don »

A most excellent engine!

Is there any way to get a linux 64 bit version? There are Pascal compilers for Linux, probably the same one you are using.

If you know nothing about Linux, I could help you get started - you can install Linux on a USB drive, as a dual boot (without hurting your existing windows install) or even just run the live CD. I could give you ssh access to a linux machine and probably several other people here could.
rvida wrote:Download link: http://www.mediafire.com/?2v80239zxj82yud

Book support

In addition to its own .cbk book format, Critter can now read Polyglot .bin books too. Just set the "Book File" UCI option to point to a *.bin file. Book creation from pgn file ('makebook' and 'mergebook' console commands) works only with native format.

Repetition detection

Critter now scores 2-fold repetition as draw. See this thread on rybka forum.

Enjoy

Richard