Measuring the randomness of online shuffles (esp with regard 2 bridge)

[partygirluk]

Hi,

I saw a post on another board wanting to know how to verify the randomness of an online bridge sites shuffle. They have access to many many thousands of bridge hands and some great programmers.

[Siegmund]

The things I'd be most concerned about are straightforward independence of the position of any two cards in a deal, and comparison of consecutive deals to show that a) any one card is equally likely to land in any hand next time and b) two cards in the same hand this deal land certain distances apart in the next deal with the correct probability.

Doing this last one will need done slightly differently according to whether the site reshuffles the previous order of the deck to generate the next hand, or uses a 96+bit RNG to select a permutation of a standard starting position.

In practice, the biggest concern in duplicate bridge is making sure the same hands don't get used in two separate events. That is, properly seeding the RNG when new sets of hands are generated, and simple bookkeeping (not handing out the same hands in two events in real life, and proper security of who has and hasn't already seen a hand at an online site.)

Do you know which site he's interested in? I was a tournament director at Swan Games, and did some testing to make sure that cards dealt at practice tables weren't the same hands as in the social room or tournament room when the practice feature was added. Also have had some contact with the programmer who wrote the software used to generate secure hands for major European events.

Your friend will find perusing the material about cryptographically strong deal generation at http://www.alesia.dk/shf/design.htm interesting, even if he's not using exactly the same testing methods or facing the same deal generator.