Hello,
I asked HGM because of the playing strength of N.E.G. v1.2
http://home.hccnet.nl/h.g.muller/dwnldpage.html
He said:"Not really. Non-searching engines like that (POS and Brutus Random are other examples) are so weak that the gap with even the weakest searching engine is too large to do meaningful rating measurements. The weak searching engines are often weak because they are buggy. This makes them forfeit games by crashing, which violates the rating model (as the probability they crash is virtually independent of opponent rating, and they could lose against arbitrary weak opponents by that.
One of the things I would like to do is to create a series of engines to fill the gap between a random mover and very weak searching engines."
Perhaps a survey of "random mover" chess programs will be created by CCWiki. Any suggestions or examples for such engines?
Best wishes,
Norbert
"random mover" chess programs
Moderators: hgm, Rebel, chrisw
-
- Posts: 1643
- Joined: Tue May 20, 2008 4:57 pm
- Location: Augsburg - Germany
-
- Posts: 27790
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: "random mover" chess programs
My plan is to make an engine that would play by meansof the Beal effect, where random evaluation would create a statistical preference for positions with large mobilty, and thus likely strong material. Increasing search depth should then increase the strength.
The problem is how to handle checkmate. If you detect that in the normal way, and assign an overwhelmingly large score to it, it highly unbalances the thing: you get an engine that plays like a complete idiot, and then suddenly announces mate in 8, which it plays out perfectly. Of course in playing against each other the engine that recognized accidental mate threats from a larger distance will achieve a better result, but it would not have anything to do with the Beal effect.
To get a more balanced way of treating checkmates, I want tomake it play according to the following rules: exposing a King to check is legal, and the game continues after King capture. But a side that does not have a King can only play null moves. This makes it very likely that capturing a King will get a high reward, the higher the earlier you capture it. As the side capturing the King will continue to expand the tree, and can pick the leave with the largest random eval without the opponent being able to steer him away from it.
Only stalemate (and other draw conditions like 3-fold rep and 50-move) would have to be recognized as game end (and receive a random score).
The problem is how to handle checkmate. If you detect that in the normal way, and assign an overwhelmingly large score to it, it highly unbalances the thing: you get an engine that plays like a complete idiot, and then suddenly announces mate in 8, which it plays out perfectly. Of course in playing against each other the engine that recognized accidental mate threats from a larger distance will achieve a better result, but it would not have anything to do with the Beal effect.
To get a more balanced way of treating checkmates, I want tomake it play according to the following rules: exposing a King to check is legal, and the game continues after King capture. But a side that does not have a King can only play null moves. This makes it very likely that capturing a King will get a high reward, the higher the earlier you capture it. As the side capturing the King will continue to expand the tree, and can pick the leave with the largest random eval without the opponent being able to steer him away from it.
Only stalemate (and other draw conditions like 3-fold rep and 50-move) would have to be recognized as game end (and receive a random score).
Re: "random mover" chess programs
Just to be clear: pos is not a random mover.
QueenBee more or less is as it does monte carlo.
QueenBee more or less is as it does monte carlo.
-
- Posts: 2204
- Joined: Sat Jan 18, 2014 10:24 am
- Location: Andorra
Re: "random mover" chess programs
I have done a version of Andscacs that can move random:
www.andscacs.com/andscacs_r087007.zip
Seting Random to 10000 will play all the moves random. Setting it for example at 153 will play 153 random moves of each 10000.
If you don't set it, it will not play random moves.
www.andscacs.com/andscacs_r087007.zip
Code: Select all
Andscacs R0.87007 by Daniel Jose
uci
id name Andscacs R0.87007
id author Daniel Jose
option name Ponder type check default false
option name Hash type spin default 128 min 1 max 16384
option name Clear Hash type button
option name NullMove type check default true
option name MultiPV type spin default 1 min 1 max 100
option name Threads type spin default 1 min 1 max 32
option name AlwaysFullPv type check default false
option name Random type spin default 1 min 1 max 10000
If you don't set it, it will not play random moves.
Daniel José - http://www.andscacs.com
-
- Posts: 5228
- Joined: Thu Mar 09, 2006 9:40 am
- Full name: Vincent Lejeune
Re: "random mover" chess programs
I think it's impossible to rate random engines because they lose all (over 999 999 on 1 000 000) their games even against the weakest engines.
Even winning a position with KQK is very unkilely ...
An amazing experience would be : a random engine vs a weak engine with only K+Q
Even winning a position with KQK is very unkilely ...
An amazing experience would be : a random engine vs a weak engine with only K+Q
Re: "random mover" chess programs
Oh I also created an engine called "ParisHilton" which does a negate on the evaluation score. Also very interesting game play
-
- Posts: 27790
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: "random mover" chess programs
N.E.G. is not really a random mover. The proper title of this thread should have been 'non-searching engines'. N.E.G. decides what tomove basedon the number of attackers and defenders of asquare, and the value of the piece on it. And whether the move delivers check. It once won a game in the ChessWar promo competition, becauseit preferrs save checking moves over non-checking moves, and the opponent was unwise enough to expose himself to fool's mate.Vinvin wrote:I think it's impossible to rate random engines because they lose all (over 999 999 on 1 000 000) their games even against the weakest engines.
Even winning a position with KQK is very unkilely ...
An amazing experience would be : a random engine vs a weak engine with only K+Q
-
- Posts: 180
- Joined: Mon Sep 03, 2007 9:15 am
Re: "random mover" chess programs
Arminius already has a command line option --Rnd that makes it a random mover.
It's easy for me to make another random mover that always mates in 1 if possible, but otherwise plays random moves.
It's easy for me to make another random mover that always mates in 1 if possible, but otherwise plays random moves.
-
- Posts: 46
- Joined: Sun Nov 30, 2014 12:06 am
Re: "random mover" chess programs
CCRL 40/4 has a random move maker BrutusRND. It is placed last, as you could expect, but, surprisingly, it has a positive rating - 210.
http://www.computerchess.org.uk/ccrl/40 ... t_all.html
I have played several games against it trying to deliberately lose to it. Mostly it has managed to avoid stalemates with the bare king against almost an intact army. Looks like the rating difference between the worst possible play and a random mover is ca 200 elo. I don't know if it is conicidental...
http://www.computerchess.org.uk/ccrl/40 ... t_all.html
I have played several games against it trying to deliberately lose to it. Mostly it has managed to avoid stalemates with the bare king against almost an intact army. Looks like the rating difference between the worst possible play and a random mover is ca 200 elo. I don't know if it is conicidental...
-
- Posts: 3226
- Joined: Wed May 06, 2009 10:31 pm
- Location: Fuquay-Varina, North Carolina
Re: "random mover" chess programs
BrutusRND achieved 64 draws, most of which were due to flaws with its opponents. Several opponents could not checkmate even with an overwhelming material advantage.nimh wrote:CCRL 40/4 has a random move maker BrutusRND. It is placed last, as you could expect, but, surprisingly, it has a positive rating - 210.
http://www.computerchess.org.uk/ccrl/40 ... t_all.html
I have played several games against it trying to deliberately lose to it. Mostly it has managed to avoid stalemates with the bare king against almost an intact army. Looks like the rating difference between the worst possible play and a random mover is ca 200 elo. I don't know if it is conicidental...
http://www.computerchess.org.uk/ccrl/40 ... Brutus_RND