I agree, but as a first pass assuming the symbols are equally weighted and there is no correlation between reels seems like a reasonable assumption. It should be true in video poker and I think it is in a fair number of other slots, like jackpot party. To check for unequal symbol weighting and correlation between reels would require a larger sample set, but we're talking thousands or tens of thousands, not millions. In video poker the reels have 52 symbols but repeats are excluded (that would be one correlation that should never happen, if it does time to snap a pic and call gaming) so the possible correlations between any two reels are 52x51-2652. Here's an experiment: make up a deck of 4 Aces, 1 King and 1 Queen, then forget what the composition is. Start drawing one card and then reshuffle and repeat, recording your results. How many samples do you need before you begin to think there are multiple aces in the deck? How many samples do you need to begin to make a reasonable guess at how many extra aces are in the deck?
AAAKQAAKAAAQA...
9:2:2
--- In vpFREE@yahoogroups.com, "vp_wiz" <harry.porter@...> wrote:
>
> nightoftheiguana2000 wrote:
> >
> > On a related note, here is the Wizard's description of how he took some random slot results (212 spins) and figured out the game with them:
> > http://wizardofodds.com/games/slots/jackpot-party/
> >
> > You can apply the same concept to video poker
>
>
> Well, within some constraints. The key constraint is that Michael presumed for the slot game that the symbol distribution was random. Piece of cake, once you make that presumption.
>
> When you start questioning the randomness (the initial subject you raised in this thread), evaluating expected return becomes much more complex. (For example, just because symbols occur on each reel with an observed frequency doesn't mean that the reels operate independently of each other).
>
> Ditto, when it comes to vp evaluation when fairness comes into question.
>
