Computer scientists have made many seminal breakthroughs while solving games. When it's a game of complete information, brute force can be applied. Draughts is solved. Chess and backgammon programs outplay the best humans. Go is the last frontier. The Asian territorial game's 19x19 matrix leads to too many legal positions for brute force. Thus far, grey cells beat silicon.
Computers are also good at incomplete info games with an emphasis on pure number-crunching. They win at Blackjack, etc. But in bridge and poker, the best humans win since both need communication skills. Poker players must figure out when competitors are bluffing, mask their own intentions, and work out the odds. Bridgeplayers must exchange info with partners and decode info exchanged between opponents and also work out cardplay.
Jeopardy is orders more difficult in terms of communication skills though it is a complete info game, given search engines. Asking the right question, for a given answer, requires offbeat retrograde analysis. There's often multiple choices and a large psychological component in formulating the most likely question.
IBM's new monster “Watson” showed recently that it can beat the best humans at Jeopardy when it outgunned champions, Ken Jennings and Brad Rutter over a three game series. Watson has about 100 x the processing power of Deep Blue (1997). This was truly impressive though Watson's occasional glitches showed it is light years from Douglas Adams' ideal (“42”!).
In figuring out how to program chess analysis, many clever short cuts and pruning techniques have been developed. The routines developed to teach Watson Jeopardy are likely to be adapted to medical diagnostics and consumer market research software.
It's unknown if Topalov will have access to the Bulgarian super computer he used for the 2010 title match, while preparing for the 2011 Candidates. If he does, the Rybka team and Vasik Rajlich will probably be involved..However, Rybka's domination of chess programming is being challenged by Houdini, a free engine written by Belgian programmer Robert Houdart. Check out version 1.5a which is backed by the Gaviota Tablebases. It's outstanding. There's a lot of needle between the Bulgarians and Vlad Kramnik in particular. So the Russians could pull out the stops in terms of computing teams as well.
The diagram, (Gabuzyan Vs Belous, Aeroflot 2011), Black to Play, displays the computational talent of the untitled teenager who won the Moscow Open. Black's solution was very elegant 25...Ne2+! 26.Nxe2 Rxg2+! 27.Kxg2 Qxe4+ 28.f3 Qxe2+ 29.Rf2 Rxf3!! 30.Rxe2 Rxc3+ 31.Kh2 Ra3 32.Rf2 Nf5 33.Rb2 Kf7 (0-1). Upon silicon analysis, it seems that 25.— Nf3+ was an alternate win since 26. Bxf3 Rxf3 27. Qxf3 Bxe4 28 Qc3 Qxh3 or 28. Qe3 Bd5! is clear enough .
