Knowledge about the players, what shape they are in, etc. are by far the most important thing. Statistics won't help you (almost) at all.
Good post Comet.
How would this be better than saying that if player X wins 80% of the games, he should win the set 4-1?
The only trick I see is the observation that 4-1 gives you two chances at 1 point (5-0, 3-2) while 5-0 only gives you one chance (4-1). So sometimes it might be worth it to pick 4-1 even though the probability is slightly higher for 5-0 (if there is a chance for 3-2). But... This doesn't make much difference in total...
Good post Comet.
I would create a normal distibution for each set, then calculate the mean and the standard deviation,(correct for discrete realisations) and then aggregate it to one distribution
How would this be better than saying that if player X wins 80% of the games, he should win the set 4-1?
The only trick I see is the observation that 4-1 gives you two chances at 1 point (5-0, 3-2) while 5-0 only gives you one chance (4-1). So sometimes it might be worth it to pick 4-1 even though the probability is slightly higher for 5-0 (if there is a chance for 3-2). But... This doesn't make much difference in total...