Ahh.. hAl:
100% chance will of course always be the least risky, but it is also a little costier.
I think his point however is that 50% chance is better than 90% chance. This is because of two reasons:
1.The short term un-randomness often experienced with PA-'random' number generater. (which may be an urban myth). You are less suspectible to this when doing larger numbers
2. The variance actually *is* lower when comparing the two. The high scan alternative is both higher on the central probabilities, and higher on the extremes.
|