Say I'm flipping coins and betting £1 on each flip. After 100 flips with infinite money, I have an 18.4% chance of being up by £10 or more (binomial with p = 0.5, n = 100, x = 55).

If I start with a finite amount of money, say £5, and cannot continue if I go broke, presumably my probability of finishing up £10 or more is less.

How would a statistician factor this into the model, to get a better estimate of probability of winning at least £X?

  • $\begingroup$ What is the "x = 55" in the first paragraph? $\endgroup$ Oct 9 '12 at 1:12
  • 2
    $\begingroup$ @gung number of successes; we require P(X >= x) $\endgroup$ Oct 9 '12 at 1:33
  • $\begingroup$ If you continue until you reach £15 (up £10) or £0 (down £5), and then stop, then the probability of reaching £15 first is $\frac{5}{5+10} = \frac13$, provable by a simple induction. $\endgroup$
    – Henry
    Oct 9 '12 at 7:12
  • $\begingroup$ @gung: $55 = 10 + \frac{100-10}{2}$ $\endgroup$
    – Henry
    Oct 9 '12 at 7:26

Personally I would calculate the recursion, with $$P(X_{n+1}=x)= \frac12 P(X_{n}=x-1) + \frac12P(X_n=x+1)$$ but with $P(X_{n+1}=0)= P(X_{n}=0) + \frac12P(X_n=1)$ and $P(X_{n+1}=1)= \frac12 P(X_{n}=2)$, and starting at $P(X_{0}=5)=1$.

Then as far as I can see, $$P(X_{100} \ge 15)= \sum_{x=15}^{105} P(X_{100} =x)\approx 0.1556568.$$

You also have $P(X_{100} = 0) \approx 0.6172994.$

  • $\begingroup$ Thanks. Is there a common software toolset for such problems, or would a professional statistician use whatever programming language they were familiar with? $\endgroup$ Oct 9 '12 at 15:06
  • $\begingroup$ @Luigi: I used MS Excel for the numbers above. $\endgroup$
    – Henry
    Oct 9 '12 at 19:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.