Steve Fossett and My Baye's Coincidence Theory

[David Sklansky]

The prior probabilities could easily be something like this.

1. 1 in 30,000 chance anyone would stage their disappearance.

2. Given Sklansky's insight into Fossett psychology, boost it to 1 in 10,000 chance Fossett would stage his disappearance.

3. Given that Fossett would stage his disappearance, 1 in 10,000 chance he would stage it in this particular way at this particular time.

4. So the parlay for the prior probability that Fossett was going to stage his disappearance in this rather unspectacular way, 1 in 100 million.

5. 1 in 1000 chance of Fossett's plane crashing in the flying conditions that day.

6. 1 chance in 5 the crashed plane would not be found after search to date.

So with these prior probabilities, Bayes' Theorum gives the odds against Fossett staging his disappearance with this flight at about 20,000 to 1.

Maybe the true odds are closer to 1000-1. Who knows? These prior probabilities are all guess work anyway. The better method would be to look at all the planes that have disappeared and gone unfound after considerable search. How many were later found and how many turned out to be cases of staged disappearences? Who has better access to such information? The experts who gave their opinion based on years of experience or David Sklansky stupidly applying logic and probability to his guesswork.

PairTheBoard

Firstly those who gave their opinion were NOT using this information. Secondly your method won't work in this case because the sample size is so small- Expert attention seeking aviators whose plane can't be found in a small area after a thorough search.

As for your #s 4-6

The parlay is not 10,000 times 10,000 but rather 10,000 times the number of times he does things like take off in a plane. Maybe five million.

The chances he will crash is nowhere near one in 1000. Worst case is one in 25,OOO.

I'm assuming that not finding the plane is about one in ten. That could be way off.

So that would make the final answer 20-1. IF it started out as a 10,000 to one shot that he would do something like that. I'd say its 2000-1.

BUT NONE OF THIS REALLY MATTERS. All your numbers could be right but you still seem to be missing the point. Which is that most people who supposedly have expert opinions on this sort of thing are far mor inexpert than either you or I because they don't have a clue how to adjust their probability estimates when evidence outside their particular field comes into play. Just like those poker players who when moved in on, don't adjust to games where the nuts are harder or easier to come by.