[Transference]

a) how often does suspect call a bet when he cannot possibly win the hand. E.g. the suspect put money in the pot when he could not win x out of y times (z%).
b) how often does suspect bet or raise when he currently has the best hand. E.g. the suspect added money to the pot when ahead x out of y times (z%).

A note on 'proof.'
I think previous posters are on the money when suggesting that a statistical representation of how unlikely the given outcomes can be is pretty important. I feel like a mere random chance expectation is adequate given how vast the discrepancy should be. Alternatively a 60/40 analysis could be done to adjust somewhat for the effects of skill. Eg. If an opponent consistently folded with the worst hand 60% of the time the odds of him folding the worst hand 45 out of 50 times would be 1 in x.