So you say a manager should intentionally bet against the odds? Those mistakes in judgement accumulate over a season whether it's the same pitcher/hitter or different ones, no different.
I say a manager should stick with a pitcher who is settled. That was Poulin. The betting odd make sense over many tries, but not for each seperate throw. For instance, we remember when Matt Williams pulled Jordan Zimmermann in Game 2 of that playoof game. For another example, I remembered the Braves-Yankees World Series of 1957. In Game 1, Mickey Mantle went 2-for-4 against Warren Spahn. Both singles. In Game 4, Mantle went 0-for-5 against Spahn.
An individua; at-bat is not the same as a season's worth of at-bats.
In his greatest season, Ted Williams failed 6-of-10 times. In the Nats' worst recent seasons, they still won about 40% of games.
Analytics cannot tell you what to do in any given spot in any inning.