0
$\begingroup$

Problem statement:

Two players play the following game of several rounds. If some player wins a round, the winner is given $1$ point, and loser is given $0$ points, and there are no draws. The game ends when some player gets $a$ points.

It is known that probability that first player wins a round is $p$. Find probability that difference between winner's and loser's score is equal to $b$.

My approach:

First, notice that if the difference between winner's and loser's score is equal to $b$, then the final scores must be $(a, a - b)$ or $(a - b, a)$.

Suppose that the final scores are $(a, a - b)$. Consider an integer lattice with points $(0, 0)$ and $(a, a - b)$. Number of paths from $(0, 0)$ to $(a, a - b)$ such that we move only up or right is $\binom{2a - b}{a}$. Now suppose that probability of "up"-movement is $p$, then the probability of "right"-movement is $1 - p$. Hence, because any such path consists of $a - b$ "up"-movements and $a$ "right"-movements, the probability for such final scores is $$p_{(a, a - b)} = \binom{2a - b}{a} (1-p)^{a} p^{a - b} .$$ Similarly, the probability for final scores $(a - b, a)$ is $$p_{(a - b, a)} = \binom{2a - b}{a - b} (1-p)^{a - b} p^{a} .$$

Hence, the final probability: $$ p' = p_{(a, a - b)} + p_{(a - b, a)} = \binom{2a - b}{a} (1-p)^{a} p^{a - b} + \binom{2a - b}{a - b} (1-p)^{a - b} p^{a} .$$

Is my approach correct? I've tried to simulate the game in Python and got different result ($\approx 0.06$ by the simulation vs. $\approx 0.105$ by the formula for $a = 11, b = 3, p = 0.3$).

$\endgroup$
1
$\begingroup$

It seems that your calculation includes the possibility, for example, that one player wins $a$ times and then the other player wins $a-b$ times, since this would be among the counted paths from $(0,0)$ to $(a,a-b)$, right?

Edit: To fix the calculation, I think you should consider all paths from $(0,0)$ to $(a-1, a-b)$, and then multiply by $p$ (and similarly $(a-b, a-1)$, and $1-p$).

$\endgroup$
1
  • $\begingroup$ Good catch. Do you see how to fix the calculation? $\endgroup$
    – saulspatz
    Dec 6 '18 at 22:57

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.