Quote:
Originally Posted by pk_nuts
Its almost a blue print to beat the casino. How it is on the interweb is beyond me????
Biased wheels are real, but the problem is to figure out where the positive biases are. Casinos routinely change wheels to prevent this sort of problem; if you change the wheels faster than an accurate data set can be collected, you can have horrendously biased wheels so long as nobody knows exactly how they're biased.
Even with an unbiased wheel, 20,000 spins isn't enough to guarantee long-term results; someone betting 1 number for 20,000 spins on an unbiased wheel will win between -$2,700 and +$600; you actually need about 50,000 spins before someone's a guaranteed loser and about f^2 that (where f is the fraction of the true odds that is the bias) before you can mathematically prove you've got a biased number. That is, if you have a bias which is either 2x or 0.5x the normal, you need about 200,000 spins to prove it; if you have a bias which is 1.1x or 0.9x normal, you need about 5,000,000 spins to prove it. (For reference, at a traditional roulette game at a live casino, at 20 spins/hr, it would take about 6 years to go through 1,000,000 spins.)
Things would be a little different if there were some SYSTEMATIC flaw in the wheel, like one whole side of the wheel is biased over the other. The reason this would be different is because the variance for betting large sections of numbers is considerably lower than the variance for betting a single number. As a result, with just a 200*f^2 spins, you could easily prove a 18-square bias, or with 4000*f^2, you could prove a 9-square (quarter wheel) bias.