A and B play until one scores 2 points in a row, who will win. Probability of A or B scoring a point is $a$ or $b$, respectively. What is the probability that A wins?
I know the correct answer is $\frac{a^2}{1-2ab}$, based on this.
I was wondering what is wrong with the following approach. We form the following tree. The green/red rectangles are where A/B has won.
Tree showing possible sequences leading to the winning of A.
We can see the following pattern for the top and bottom rows (green boxes where A has won):
Top: $a^2, a^3b, a^4b^2, \ldots$
Bottom: $a^2b, a^3b^2, a^4b^3, \ldots$
When we add the probabilities, both top and bottom rows are geometric series with a factor of $ab$. So,
Sum of the top row: $\frac{a^2}{1-ab}$
Sum of the bottom row: $\frac{a^2b}{1-ab}$
Total probability: $\frac{a^2(1+b)}{1-ab}$
Questions: This not equal to $\frac{a^2}{1-2ab}$. What am I missing?
Your game is not the same as the game in the linked question.
I did not look at your infinite tree. Anyway: The result is correct.
Apart from the initial state $O$ my graph (it's not a tree) has just two nonterminal states: last game won by $A$, and last game won by $B$. Denote by $p|_A$, $p|_B$, and $p|_O$ the probabilities that $A$ wins when we are in state $A$, $B$, and $O$ respectively. Then we have the equations $$p|_A=a\cdot 1+ b\cdot p|_B\>,\qquad p|_B= a \cdot p|_A+ b\cdot 0\ ,$$ resulting in $$p|_A={a\over 1-ab},\qquad p|_B={a^2\over 1-ab}\ .$$ It follows that $$p|_O=a\cdot p|_A+b\cdot p|_B={a^2(1+b)\over 1-ab}\ .$$