Assume that we have a biased coin with probability $p_1$ of getting H and $1−p_1$ of getting T on the first trial, $p_2$ of getting H and $1−p_2$ of getting T on the second trial and so on such that $2/3<p_1<p_2<p_3...<p_n<1$. The probability $p_i$ of getting H increases as long as we get head in a row. If a tail appears, then we reset to probability $p_1$ of getting H in the next trail and so on.

What is the expected number of trials to get $n$ H in a row?

  • $\begingroup$ You are using $n$ twice in the question, once to determine the 'end' of the sequence of $p_i$ and, again when you are setting the desired number of consequtive $H$ to complete the experiment. Is this intentional? Since $p_i$ are bounded by 1, it's natural that the sequence must be larger than the number of heads to complete experiment $\endgroup$ – jameselmore May 3 '15 at 20:22
  • $\begingroup$ Yes, because if we get T we restart the trials with $p_1$, the experiment ends when we get $n$ in a row. $\endgroup$ – Ahmed Younes May 3 '15 at 20:30
  • $\begingroup$ Ah, and you ignore previous trials if a $T$ is thrown? $\endgroup$ – jameselmore May 3 '15 at 20:33
  • $\begingroup$ I am not sure but should be done about that, suggestions are appreciated. $\endgroup$ – Ahmed Younes May 3 '15 at 20:37

I'll use $S_i$ to denote the state where we have $i$ heads in a row, $N_i$ the expected number of trials to get to this state for the first time.

To get to the $S_{i+1}$, you first need to get to $S_i$ then you toss the coin. If you get $H$, then the number of trials is $N_i+1$. If you get $T$, you get back to $S_0$ and the number of trials is $N_i+1+N_{i+1}.$

So we have:

$$N_0=0$$ $$N_{i+1}=(N_i + 1)p_{i+1} + (N_i+N_{i+1}+1)(1-p_{i+1})$$

This gives us:

$$\begin{align} N_{i+1}&=\frac{N_i+1}{p_{i+1}} \\ \end{align} $$


$$\begin{align} N_0&=0 \\ N_1&=\frac{1}{p_1} \\ N_2&=\frac{\frac{1}{p_1}+1}{p_2}=\frac{1}{p_1}+\frac{1}{p_1p_2} \\ &\ \ \vdots\\ N_n&=\frac{1}{p_1}+\dots +\frac{1}{p_1\dots p_n} \end{align} $$

If I've made a mistake, please let me know.


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.