It's not the degree of deviation which tells you a wheel might be exploitable but the correlation it has with some parameter.

No, the deviation from the norm...the standard deviation can and often times will indeed tell you that a number or section is indeed abnormal and biased.

Correlations are also for comparing one spin sample to another. Yes, you can also say "correlating" a possible defect with what appears to be showing as bias in your data. This type of observation and skill is very very valuable. For example, looking at the light reflection movements off of various components of the wheel enabled me to nearly instantly determine which wheels were biased and worthy of tracking, and which wheels to ignore, and enabled my team and I to find incredible success. Be careful though. If you look for defects, then you'll likely find them. Unfortunately most of them are insignificant and other big defects that may have a real effect can be completely overlooked. Not all defects will have a local effect... where the defect is viewed, but may have a resulting effect on numbers that are far away from the actual defect. Not all defects can be seem. By the way, don't just look for defects, listen for them as well.

Applying creative observation, some engineering skills, and a little math can take you a long ways in this game.

I don't think these days it's possible to find a biased wheel just by collecting numbers,

Yes, it's definitely still possible. However, break the tracking down into more parameters/conditions and you'll find much better pay dirt.

but if the degree of *correlation* is strong then it suggests causality although you have to be careful because correlation is not causation. The deviation might only be small but if it is consistently associated with some external parameter then you may be onto something.

If the deviation is small, then you must track whatever it is that you're testing for more trials...after which the sd should grow larger. The more trials you test, the larger the standard deviation should grow. As should the chi square.

What causes hot numbers?

Hot numbers...it's usually just randomness. Biased numbers though are something entirely different. What causes bias? Here's the short list:

1. Loose components related to the pocket compartment, frets, or pocket inserts.

2. Piss poor assembly of wheel components.

3. New wheel feature gimmicks that malfunction or that aren't engineered very well. (Love this one)

4. Ridges in the outer number tape of the wheel plaques.

5. Warped domes/cones

6. Rise and fall of poorly assembled/bolted indexable and nonindexable pocket compartments

7. Wheel wobble from bad bearings.

8. Some types of debris...including spilled beer.

9. Damage/wear