Usually, we perform chi square test on a wheel to see how much it deviates from a random wheel.
Let’s say we know that a wheel is not random and bias enough to have a good advantage. Can we assume the distribution of the data we took of a wheel, then use data of preprofiling (the 200 to 400 spins of data we take before actually playing) to see if it lies between a small chi square number instead comparing with the data took before, perhaps over 5000spins? Or this is redundant?
Let’s say instead of assuming each number is 1/37 on a unbiased wheel. However, after taking data, we found out some numbers are 1/30, some are 1/32, some 1/42, etc, can we do statistcal interference with this instead?
I can definiely defect spot and tell if it has the same rotor. However, i cant tell if conditions change too much so needs to use statistical test to tell and avoid playing, such as cleaniness, storm front, ball damage, etc.
Especially for wheel head wobble, conditions can change very fast, probably due to air pressure and the range of numbers can shift 2-3 pockets perhaps in a 7hours sesson.