The Alarmed Centipede

     In the book Grandmasters of Chess Harold Schonberg wrote of Petrosian, “As a young man he played like an old one…Petrosian had antennae that could sense danger, he was like an alarmed centipede who would scuttle for the dark and lurk there, dangerous, ready to bite when threatened. But it was not going to be he who would initiate the attack.”
     Side bar: I didn’t know a lot about centipedes, so here goes: They have small mouths and have large, claw-like structures that contain a venom gland. Because most centipedes are carnivorous creatures that forage for food at night, they use their claws to paralyze their victims, such as worms, spiders and small vertebrates. Adult centipedes hide in moist, dark and secluded areas during winter. Centipedes detect prey through the use of their antennae and prey is immobilized by venom injected from their fangs. Centipedes are venomous. Their venom allows them to attack prey and defend themselves against predators and other natural enemies. Centipede venom is not normally life endangering to humans, although the bite can be painful. Back to Petrosian…
     In the game Petrosian vs. Bisguier in the Stockholm Interzonal in 1962, Petrosian made a double N sacrifice and Bisguier saw that his Q was going to be trapped. When he realized what was happening, Bisguier concluded that in order to save the Q he would end up with Petrosian having 3 Pawns for a N and a superior position which he would likely win. As a result Bisguier decided to let Petrosian have his Q for three pieces and two Pawns. They reached the position below and without making his move Petrosian offered a draw which was accepted. Bisguier wrote that white’s position appears to be somewhat better after 23.h4 and that his only plan was to keep his pieces active and react to whatever Petrosian played, believing he had good fighting chances.
     Discussing the game later, Petrosian told Bisguier that the decision to give up the Q shocked him and, not liking complications, he offered the draw. Petrosian also added that Tahl and Keres were watching the game and by not making the N sacrifices it might have made him look like a sissy to them. At the same time he was afraid he might lose so he offered the draw. Both Stockfish and Houdini think White is about 1.00 Pawn better, but I was interested to see the Monte Carlo analysis results and that’s sort of what this post is really about...Monte Carlo analysis.
     I was tinkering with my Fritz 12 program the other day and noticed the “Monte Carlo” analysis function again. I had never used it because it only works with Rybka engines which I never had. However, have purchasing ChessOK Aquarium some time back one of its engines was Deep Rybka 4 w32, so I thought I would go back and see how the Monte Carlo Method works.
     Apparently in the financial world you can’t always make precise judgments of situations so the Monte Carlo method is used to get a feel for different circumstances. Actually this method can be used in any complex situation where precise analysis is impossible…say an unclear chess position and this is what the Monte Carlo method tries to do. Monte Carlo plays thousands of ultra-fast games in a few minutes in a given position and gives result statistics. It’s supposed to be useful in endgames because it recognizes fortresses and other situations where no progress can be made and it’s also supposed to be good in Rook endings. It’s also good in evaluating exchange sacrifices or other types of sacrifices that have been made for positional compensation. Using this method of analysis you’ll know the moves that give you the best odds. It’s not good for tactical situations though, but rather those where positional judgment is called for. Another situation where it might be good is a position that doesn’t have a lot of theory.
     Sometimes the real truth of a position may not be apparent for many, many moves and in the infinite analysis mode the engines may not be able to see that far ahead, but if it plays enough games from the starting position a trend might develop. Another thing it does is it stores its analysis (if you want to save it) in a tree formation, so this could be handy for some opening positions. The big question is of what practical use is this information? I entered a Rapid tournament on LSS and being intrigued by the risky (unsound) Urusov Gambit (1.e4 e5 2.Bc4 Nf6 3.d4) I played it in two games and used the Monte Carlo method in conjunction with the usual engine analysis. The result was I played a lot of opening moves that were not in the engine’s top several choices and ended up with positions that give better than even winning chances. In one position the results were based on nearly 69,000 games!! The games are still ongoing, so I will have to hold my final opinion of the method in abeyance.

In this position I let it play 2047 games with the following results:
23.Rc3 was played in 1779 games and white won 47% lost 19% and drew 34%
23.Rc7 was played in 135 games and White won 62% lost 18% and drew 20%
23.h4 as suggested by Bisguier was played in 133 games and resulted in white winning 38% losing 25% and drawing 37%

Based on this it appears that white’s best move is 23.Rc7 as suggested by SF5 and H2, but as I experienced in the Urusov Gambit games, it does not always work out that the engine's top suggestion gives the best results. In any case, had he played on Petrosian most likely would have won. I am curious if any readers have experience with the Monte Carlo Method.