Finding the value of an biased coin flip game with stop loss

346 Views Asked by At

Question: I have a biased coin, with 70% H and 30% T. I receive \$1 each time I flip a H and lose \$1 for T.

I am going to flip the coin 100 times, without any constraints, the expectation of money I will received will be \$40.

Now, what is the expectation if I stop once lose \$10?

My thoughts: I model this as a random walk, $S_t = \sum_{i=1}^t Z_i$. And $\tau$ be the stopping time when $S_t$ reaches -10 for the first time. Our game stops at $\tau \wedge N$, where $N = 100$.

Now the expectation will be:

$\mathbb{E}[S_{\tau \wedge N}] = -10p + 40(1 - p)$, where $p = P(\tau\leq N)$

If the coin is unbiased, then $p$ can be found using reflection principle. Not sure how to find the distribution of $\tau$ for the biased coin. Any suggestions are welcome!

Edited: I just realized, if the coin is unbiased, then $S_{t\wedge N}$ will be a martingale and we can use optimal stoping theorem to get $\mathbb{E}[S_{\tau \wedge N}] = \mathbb{E}[S_0] \mathbb{E}[\tau \wedge N] = 0$. We can apply this theorem because $\tau \wedge N$ is also a stopping time and bounded.

1

There are 1 best solutions below

4
On

This is not a complete answer (since I did not find a closed form for this probability and even suspect it may not exist at all), but here are my thoughts that may give you a better insight to the problem.

First, for a fixed $k$ and a fixed sum $S$ let us count the following probability

$$f(k, S) = P(z_1 + z_2 + \dots + z_k = S)$$

For all $S > k$ and $S < -k$ this probability is equal to 0. Note that $z_k$ is either $1$ or $-1$ with probabilities 0.7 and 0.3 respectively. Thus,

$$f(k, S) = 0.7f(k - 1, S - 1) + 0.3f(k - 1, S + 1)$$

This formula allows computing $f(k, S)$ recursively.

Then, for some $m > 0$:

$$p_m := P(\tau < N, \tau \geqslant m) = P(\tau < N, \tau \geqslant m | \tau = m)P(\tau = m) + P(\tau < N, \tau \geqslant m | \tau \geqslant m + 1)P(\tau \geqslant m + 1) = P(\tau < N | \tau = m)P(\tau = m) + P(\tau < N | \tau \geqslant m + 1)P(\tau \geqslant m + 1) = f(m, -10) + P(\tau < N, \tau \geqslant m + 1)(1-f(m, -10)) = \\ = f(m, -10) + (1-f(m, -10))p_{m+1}$$ As you can see, $p_m$ also can be counted recursively (but the closed form for $p_m$ can hardly be computed). The answer to your question is $p_1$.