That's the key point. I don't think you play for best odds of winning the next two, you play for best odds of winning game 4 because if you don't win game 4, game 5 doesn't matter. Pitching Kershaw gives them the best odds of winning today. If they win today then they can worry about Thursday. Roberts will be in Showalter's boat if he loses game 4 while his best pitcher sits on the bench.
I'll never be convinced that that argument is rational. That's just managing with your heart and cliches. You have go with the option that gives you the best chance to win the series or win the game, not the best chance to extend the series or extend the game and then hope for the best.
I suppose the argument is a little more sound if you think that Kershaw is going to be just as good today as he would be on Thursday, but I don't think any of the data includes pitching on 3 days rest when he didn't play in July and August, and is still ramping up his pitch count.
