Originally posted on 03/23/2024:
I don't think players' scores are statistically independent (only in such a case probabilities can be multiplied). Furthermore, lines should not be taken as probability inputs in a model, as they do not represent probabilities (it's just a crowd wisdom effect that most often they are pretty close to the probabilities). A sharp bookie's closing line might be taken as probability, but with removed margin (i.e. probability decreases, as bookies always suggest overall probability > 1 due to the margin).