Page 1 of 2

A better way to check engine's strength

Posted: Sun Jul 04, 2010 1:55 am
by Hagen
I've been hearing and reading people's assessments on engine strengths based on computer shootouts between different engines using opening books. Maybe a better way to check the strength of engines would be to run purely FRC tournaments between these same engines to test their ability in unfamiliar positions. If my hunch is correct you might get surprising results in the engine games. Engines one wouldn't normally think about may end up topping the field.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 1:58 am
by hyatt
Hagen wrote:I've been hearing and reading people's assessments on engine strengths based on computer shootouts between different engines using opening books. Maybe a better way to check the strength of engines would be to run purely FRC tournaments between these same engines to test their ability in unfamiliar positions. If my hunch is correct you might get surprising results in the engine games. Engines one wouldn't normally think about may end up topping the field.
The only issue is that not all engines play real FRC. And even for some that do, they have no eval modifications that can help in some starting positions. I too don't like opening book testing as that is just another factor that influences the game before the engine knows what is going on.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 2:00 am
by Hagen
It's a pity Crafty doesn't play FRC. I've been wondering when it will.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 2:36 am
by Richard Vida
The next Critter will be FRC capable. Watching a few test games I was very surprised how well it did with unknown openings. Alas, that happened mainly due to the fact that the opponents were more clueless than Critter :)
But I am still struggling to teach it how to reasonably develop its pieces according to human standards. If it survives the first 5-10 moves it plays very strong afterwards. But often it is losing due to blocking its own bishops or deploying its knights on A/H files.

Richard

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 2:57 am
by hyatt
Hagen wrote:It's a pity Crafty doesn't play FRC. I've been wondering when it will.
I have it on my to-do list...

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 2:58 am
by Hagen
I believe FRC is the best way to see if chess engines are really strong as they claim. Now Rybka 4 is probably the strongest FRC engine there is right now...but this doesn't mean it will always stay there. Personally I believe FRC is the only way to play the game of chess without worrying about opening book knowledge. I just read another thread with the depressing conclusion that 40/120 computer chess game tests result in nearly 50% draws depending on engine strength. FRC is the only resource we have left that won't be solved by computers for a very long time...and definitely won't be solved by GM's precisely because the rules of FRC forbids chess players from choosing the opening setup. When this rule is obeyed...opening book theory in FRC becomes meaningless for human GM's. There would be no way a GM could study opening theory in FRC. But at least the GM's would be able to sleep at night without spending hours analyzing the latest lines of theory to trick the opponent. An even wilder form of FRC is available...Double FRC. Same rules as FRC but each side has a different FRC setup. This leads to 921,600 possible setups. A Carlsen, Topalov, or Kramnik would never be able to memorize such a vast number of opening setups.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 4:38 am
by Uly
For the record, Vas didn't do anything special for the FRC version, he only made the engine follow the rules, and that was it.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 11:02 am
by Hagen
I get very excited to see *any* chess engine released that can play FRC. I look forward to any such release with anticipation and joy because I am beginning to see the benefits to my tactical games by playing this form of chess. It really helps me figure out what my weaknesses are. I also have come to the conclusion the chessboard is too small for playing chess games. It becomes too easy for opponents to simply swap off chess pieces willy nilly and just head for the endgames. I think serious consideration should be made to allow the chessboard being expanded from the current 64 squares to 88 squares. An extra three rows of squares would be enough to allow either side a chance to develop their pieces and maneuver them to their optimal locations before clashing with their opponents. Right now, with only sixty four squares...it becomes very easy for opponents to trap their adversaries unwary of sudden traps...especially in FRC. One reason why FRC is different because of the setup is because the pawns suddenly become far more valuable in this form of the game. The pawns can trap pieces very quickly here because once you move a piece...a sudden pawn lunge can force a piece awkwardly placed to either retreat or be sacrificed alot more quickly than classical chess. So pawn storms are very frequent in FRC. Another trick I've seen is sometimes there are players who use just three pieces in FRC to checkmate the opponent...a queen and two knights. You develop the knights quickly and move them around until you swarm the opponent's king and then sacrifice them to open up the position and win the games. These are just some of the new strategies I've seen in FRC which are virtually unknown in classical chess that suddenly becomes possible in the new form. Best thing of all about FRC...no need to buy opening books on systems like you do if you play the King's Indian, Ruy Lopez and the Sicilian.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 9:06 pm
by Uly
I enjoy studying openings, in fact, I think my opening knowledge is about 200-400 elo points over my actual level, I've beaten much stronger players just by using opening traps, and against equals it's common that I come out of the opening with dynamical advantages, a pawn, or even a knight up.

Then my real self kicks in and, if the opponent is strong enough it doesn't matter what advantages I have, he'll probably beat me, but he would have beaten me much easier if it was an even match. My opening knowledge is as if my opponent was giving me a handicap at the start of the game, or he has to play strange opening lines that I haven't studied, but that also favors me a bit.

All of that time studying openings would be wasted if I played FRC, it's as if a brick hit my head and I forgot all what I've learned, considering I studied openings until I concluded I was good enough (for my level) on that phase of the game, it'd be a huge loss, so it's out of the question.

Re: A better way to check engine's strength

Posted: Sun Jul 04, 2010 10:38 pm
by Hagen
If anything, FRC makes endgames even more interesting to play. Even GM Kasparov says the same thing but phrased his wording differently. He believed chess positions resulting from FRC are too bizarre to comment on. What I took from that remark is opening systems using FRC would force chess players to think about their strategy from the first move. This means...the endgame starts in the beginning! In classical chess in my opinion all you have are two phases...opening theory and endgames. Computers already have access to endgame tablebases and have amassed a huge amount of theory from opening analysis. So where does the originality come from? Where does it start? I think the late Bobby Fischer was right. FRC is the way to go.