Quote:
Originally Posted by HappyLuckBox
But Given a large enough sample size (infinity) wont that sequence almost surely occur? Obviously its an extreme hypothetical haha
Im probably just arguing semantics here.
If i could take 9trillion flips of that coin i obviously would do it in a heartbeat
I don't think it's all that hypothetical, and it's really nothing to do with semantics. If the way you're thinking about it were correct, no one would ever make money - that can't be, in a zero-sum game, right? Yes, that particular sequence will occur. The question is how big is your bankroll at that point?
P(E) = P(flip 1,000 tails in a row in 1,000 flips) = (0.5)^1,000 = 9.33x10^(-302)
Now how many ways can that happen within the first 101,000 flips, let's say: # Ways = 101,000 - 1,000 = 100,000 ways.
P(flip 1,000 tails in a row in first 100,000 total flips) = P(E)x(# Ways) = (9.33x10^-302)x(10^5) = 9.33x10^-298
Hope this kind of illustrates the point. Like the odds that you flip 1,000 tails in a row within trillions of trillions of flips is pretty damn good. But how ****ing enormous is your bankroll by the time it does? Like we're dealing with expectations and probabilities here, you don't automatically bust if you lose a bunch in a row.
It would take approximately 10^301 flips to get to 90% probability of having had a sequence of 1,000 tails in a row at any given point. Your bankroll after just 1% of that time had passed is 10^299[($101)(.5) + (-$1)(.5)] = $5x10^300... So probably not too concerned at that point that we've lost 1,000 times in a row. Even if a I fudged a factor of 10 or something in there, it's still pretty clearly not a concern.
Another way to think about it would be: H = number of heads, T = number of tails, SB = starting bankroll. (Remember we win $100 on heads, and lose $1 on tails) So to bust our bankroll, we'd have to reach the point where:
H($100) + T(-$1) < -$SB.
T(-$1) < -$SB - H($100)
T > H(100) + SB
Just because we reach any sequence of heads and tails at some point doesn't mean we ever satisfy the above relation. There is some probability for it (the risk-of-ruin), but it's not a guarantee.
Sorry for thread derail, btw.