Marty Murray wrote:The three game combination he is choosing is the order in which he plays against Steve and Larry. We want to see whether we can get above 51%. So we give him the greatest probability of winning by having him play Steve twice and play Larry in the middle game.
This makes intuitive sense but is actually wrong in a crucial way. Given S1, Matt should play Larry twice.
Logically speaking, we'd justify this by saying that Matt needs only to win two games
in a row. This means that the middle game is the most important: he MUST win that game to have any chance of being admitted to the snobby, twisted tennis club. So he wants the critical game to be against the weaker opponent.
Algebraically, we'd sort it out as follows. Suppose Matt's chances of beating Larry are x and Matt's chances of beating Steve are y, where y > x >Â 0.
Playing Steve twice, Matt MUST win the middle game and AT LEAST ONE of the other games.
p = x * (1 - (1 - y)²)
Playing Larry twice, Matt MUST win the middle game and AT LEAST ONE of the other games.
p = y * (1 - (1 - x)²)
Let's pretend that playing Steve twice is better. That gives us the inequality
x * (1 - (1 - y)²) > y * (1 - (1 - x)²)
or
2xy - xy² > 2xy - x²y
or
x²y > xy²
or
x > y
But this contradicts y > x.