BEST 4 CHESS ENGINES AND THEIR STRENGTHS, ANALYSES AND NOTES BY M. ERGENEKON / MS. IT ISTANBUL TECHNICAL UNIVERSITY-- HOUDINI, STOCKFISH, RYBKA, CRITTER
1) Ability to find original moves:
Houdini is the best, Rybka and Critter ara close to each other; on the short run Critter is slightly better but on the long run they are equal, maybe Rybka is even slightly better. Stockfish is very useful in middlegame and tactical positions but since it does not use tablebases it has a disadvantage against others in the endgame.
2) Position Evaluation
Here again the most realistic evaluation comes from Houdini, followed by Critter; Rybka and Stockfish. The best ranking for non-tactical evaluation is in the above order. For tactical positions and especially for middlegame combinations I suggest Stockfish.
3) The speed of anlysis
The speed in anlyses is as follows: Stockfish, Houdini, Critter and Rybka.
4) The ability to descend to depth
Again in proportion with the speed of analysis Stockfish is leading with Houdini, Critter and Rybka following. Although Rybka is the slowest engine in the speed of analysis it's ability to descend to depth is close to Critter. Here I want to express that stength of anlysis depth is not equal in all engines. Rybka's 30 depth anlysis is close to Stockfish's 40 analysis depth.
5) Does the engine get in to an infinite loop?
Houdini enters infinite loops very often. The others except Stockfish also enter infinite loops. But it seems that Houdini enters an infinite loop every time there is an unconcluded anlysis.
6) Official website
Houdini: http://www.cruxis.com/chess/houdini.htm (Commercial, author: Robert Houdart)
Rybka: http://www.rybkachess.com/index.php?auswahl=Home (Commercial, Author: Vasik Rajlich and his team), There is cluster supportfor super computers. Also Rybka has a very innovative and very strong opening book. To beat Rybka for a chess engine is a criteria for the candidacy for the world title.
Critter: http://www.vlasak.biz/critter/ (Free, closed source)
(Free and open source, Authors: Tord Romstad, Marco Costalba, Joona Kiiski)
Critter, according to its authors differ from other chess engines by choosing the Knight, in Bishop Knight exchanges. But again according to their authors in positional evaluations there is not much difference except this. They are using “singular extention” for move elimination. Also it is stated that Stockfish increases its playing power with this method.
One of the best ways to increase the the playing power of a chess engine is to enhance the algorithm called Alpha-Beta Pruning . In simple terms this means the betterment of move eliminaton method. The best engines which use this method are Critter and Rybka and maybe also Houdini but Stockfish uses (singular extention)as I mentioned above which is basicly a method where you increase the depth by one if there is a viable only move or a singular move. The other 3 engines find this method expensive in terms of using computing and computer resources. The reflection of this choice on the chess board is as follows: Stokfish becomes very strong in finding powerful moves in the middle game which lead to a win but the anlyses may miss intermezzos. But Rybka and Critter do not miss the intermezzos but since they cannot descend to depths like Stockfish they cannot come up with the conclusion to their anlyses. On the other side Houdini behaves in an optimum fashion between these two extremes and that is why it is the strongest engine.
Advice (incase you may use it!): If there is already a move in your mind and you are looking for variation instead of a move;first use Stockfish then use Houdini this will be more advantageous. But on the contrary if you are looking for a move first check with Rybka or Critter and then analyse the move you find with Houdini and Stockfish .
Among these even the Houdini 1.5a version is better than the other engines. Critter is better in (40/4 blitz) and Stockfishis much better at 40/40 (move/minute) time control games.
Stockfish's final version is very advanced but the last version of Critter doesn't look too different from 1-2 versions. (Source: http://www.computerchess.org.uk/ccrl/ )
M. ERGENEKON MS. ITU Istanbul Technical University