Chessmaster 10th edition Personality strength levels a joke?

Discussion of computer chess matches and engine tournaments.

Moderator: Ras

bigo

Chessmaster 10th edition Personality strength levels a joke?

Post by bigo »

I'm Fide rated 2050 and get my ass kicked by the 1700 rated personality in chessmaster, who are they trying to fool? One minute it's playing like a patzer and the next like Garry Kasparov. Why is it so hard for the programmers to emulate weaker play?
swami
Posts: 6662
Joined: Thu Mar 09, 2006 4:21 am

Re: Chessmaster 10th edition Personality strength levels a j

Post by swami »

bigo wrote:I'm Fide rated 2050 and get my ass kicked by the 1700 rated personality in chessmaster, who are they trying to fool? One minute it's playing like a patzer and the next like Garry Kasparov. Why is it so hard for the programmers to emulate weaker play?
I wouldn't bother with personalities of stronger chess programs. I suggest you try playing against freely available chess engines which play exactly like their given elo.
LiquidNitrogenOverclocker

Re: Chessmaster 10th edition Personality strength levels a j

Post by LiquidNitrogenOverclocker »

bigo wrote:I'm Fide rated 2050 and get my ass kicked by the 1700 rated personality in chessmaster, who are they trying to fool? One minute it's playing like a patzer and the next like Garry Kasparov. Why is it so hard for the programmers to emulate weaker play?
It's really not hard to do this. I implemented a few different "weaker" levels on my checkers program, based on ideas from Gil Dodgen.

There is one simple approach that is easy to implement, that can have a wide range of effect on the program's performance.

Add a random number to a certain percentage of the leaf node scores.

If, say, 25% of your nodes have +/- 0.1 pawns added to the score, you will get completely different principal variations.

If you increase this to 75% of the nodes having +/- 0.5 pawns added to the score, you can see how this would be a much bigger handicap, and the program will play much weaker.

So, with 2 variables, one for the %of nodes, and one for the random number range, you can make the program play weaker without making it "completely ridiculously weak", while avoiding the "Kasparov mode" as well.
JVMerlino
Posts: 1404
Joined: Wed Mar 08, 2006 10:15 pm
Location: San Francisco, California

Re: Chessmaster 10th edition Personality strength levels a j

Post by JVMerlino »

Which 1700-rated personality? It's true that the ratings for some of the personalities have a much wider error bar than others. Additionally, some personalities really do play better against humans than others. Also, as Ed has described, some CM personalities have randomness and others do not.

jm
muxecoid
Posts: 150
Joined: Sat Jan 30, 2010 10:54 am
Location: Israel

Re: Chessmaster 10th edition Personality strength levels a j

Post by muxecoid »

You must exploit the weaknesses of the specific personalities. If you know that your opponent values knight more than rooks you can put your rooks under knight's attack and your opponent will not try to win an exchange.