Frank asked:
> 1. What should one record and how?
> 2. What is a large enough sample? (And how to parse the
> sample.)
Record each type of card for each "play", for example I was dealt AdAc3h4h5h and drew 2hAhAs, that's 8 cards. Record 52 cards and you have a cycle. Record 260 cards and you have 5 cycles, which is statistically significant. At this point you are looking to verify (or disprove) that the average cycle of each card is 52. By 25 cycles you should have a fairly solid answer on whether or not any cards are missing from the deck. The next step is to look for correlation (stickyness or shuffle tracking) between cards. If the shuffle is truly random, the average cycle of each card should be 52 and there should be no correlation between cards, so for example the aces shouldn't be clumped together, or spread apart, anymore than would be predicted by probability theory.
http://wizardofodds.com/games/slots/jackpot-party/
