Adequate sample size

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Adequate sample size
Message posted by Michael Schell on November 10, 1999 at 12:00 AM (ET)

I've been taking samples of random cards cut at a popular internet gaming site (MSN Gaming Zone). I'm trying to resolve a longstanding suspicion among cribbage players there that the cards are "cooked" to encourage high-scoring, fast-paced games. I've never been convinced by this, so I've been keeping track of the cut (or "starter") card for each deal. The card is cut automatically by the computer program that "deals" the cards to each player.

I've accumulated over 1000 samples so far, and find that the standard deviation (8.31) is under 10% of the average number of cuts per card (85). Note that I am ignoring the suit of the cut card, and am simply counting the number of aces, 2s, 3s, etc. cut.

My questions are, since each of the 13 ranks should have an equal chance of being cut:

1. how large a sample set would you consider adequate to measure the true "randomness" of the cards cut by the game-playing software?

2. what standard deviation would you expect, given an adequate sample set?

Thanks for your response.

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Adequate sample size
Message posted by JG on November 11, 1999 at 12:00 AM (ET)

Since modern computers are so fast you should be able to solve the problem by 'simulation' or 're-sampling'.
Some of the available statistics sites or some Basic programming should do the job.
There is a fromula solutiion and I will post it sometimes later for you.

Re: Adequate sample size
Message posted by T on November 11, 1999 at 12:00 AM (ET)

Since 1000/13 is 77 where does the 85 come from ?
To test for true 'randomness' is a very complex process.
Getting the averages right is only a small part of the problem.
The gamling business is notorious for being 'cooked' in many different ways. There are books on the subject of cheating in gambling and there is an industry out there selling 'special' stuff to help people cheat, to help detect cheating and probably to help the establishment cheat.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.