There is a real problem when you look at the results of 6910 games
and conclude that you have found the true odds. Even worser for
Tiger's look at 1 season of Miami with 3110 games.
The problem is that they are simply not large enough samples.
If Tiger were to set his simulator to do 10 runs of 6910 games
each, with a serve advantage of say 4 or 5 percent, he might find
that the 13 exacta ranged from 33-1 or so, to about 40-1. Similar
variations would occur for doubles, when you flip and possibly
reduce the serve advantage.
However, when you crank the trial up to something huge like half
a million, you start to get very consistent results. Eg: approx
33.5-1 and 38.2-1.
Taking this further, if you examine Tiger's lead-off example of
exactas 23 and 32, you'll see he shows returns on the dollar of
.78 and 1.05 respectively. Looks intriguing. But when you jump
one step to Miami's summer season you get a dead heat of .97 each
for the same numbers.
So, even though you have an apparently large number of results,
you still have to use caution in concluding that you are onto
the "truth".
Anyway, Tiger did put in a big CAUTION to this effect in his initial post.
