A skillful blackjack player, one who counts cards, maintains some information about the distribution of cards remaining in the deck at all times.  The player adjusts both betting style and play based on this "count" information.  Depending on the rules used by a particular casino, the skillful player may have a slight edge over the casino. Without knowing exactly what the player is counting, we would like to write a program which is able to assess the player's playing skill.

There are two potential benefits from this research.  First and foremost, this is related to the much harder problem of assessing the quality of decisions people make under uncertainty.  For example, a pension fund manager tries to distinguish a good portfolio manager from a lucky one.  Second, there are many gamblers who deceive themselves into thinking they are able to play blackjack well enough to beat the casino.  In fact, casino blackjack revenues skyrocketed after Thorpe published his landmark book, Beat the Dealer, which explained how to effectively count cards.  Players who discover their true skill (usually very poor) will hopefully be deterred from gambling.  (As an aside, I suspect this sort of research is conducted by casinos who, due to their financial interests, are disinclined to
publish results in the area.)