Quote:
Originally Posted by David Sklansky
I just hope that everyone realizes that the above statement is flawed. If two thirds of green coins are weighted 70-30 heads, while other coins are 50-50, a green coin that came up heads 5 out of ten flips has an EV of at least 51 heads out of its next 100 flips. In other words the two orbits might give you "better info" but not to the point where the original info should be ignored.
Saying the heuristic isn’t useful clearly isn’t intended to claim that it would in a theoretical framework be worth nil. You’re being pedantic.
If you crunched the numbers and it turned out hispanics were 5% more likely to raise the river as a bluff, it’s not even really necessarily theoretically sound to adjust other reads by that 5% unless you think that the driver of that 5% difference is independent of other info gathered.
Ie: if Hispanic’s skew younger significantly, and younger people bluff raise more in a similar proportion to what their demographics would imply, if you add the 5% in addition to whatever figure you’re adjusting for age there’s a good chance you’re double counting and overestimating the increased frequency of bluff raises.
If you thought there was some quality in hispanics that was independent of all other variables you’re taking notice of then yes, the residual would exist. But no matter whether you think that residual exists, a lot of your initial race-based read will be eaten up by other info. And I can’t imagine anyone thinking the difference is large to begin with which is why saying that it stops being useful is a pretty fair statement.
if anyone thinks they do have strong race based reads though, again I’ll say, make a specific prediction and we can find a way to bet on it.