Originally Posted by Rojo
Somebody here had the Tommy Lasorda tag about "the best teams loose a third of their games, the worst win a third so it's the middle third you have to worry about".
Games decided in the late innings are marginal. But the baseball gods are commies, they want to push everyone to .500. ergo, the marginal becomes crucial.
And yet, every year, the "ability" of a team to win 1 run games proves to be more or less random.
Lasorda's characterization is fun, but really misrepresents what's actually happening. The trick is realizing that games aren't first put in to a bucket called "1 run games" and then
decided based on a team's ability in those games. Games are decided by the accumulation of runs over the course of a game. A run in the 3rd counts just as much toward the final outcome as a run in the 9th. What matters is how many more of them you can accumulate than your opponent.
And on the season level, teams improve by shifting their whole distribution of runs scored up relative to their distribution of runs allowed, essentially taking games from the definite loss bucket and pushing those in to the "margin" while taken games formally in the "margin" bucket and pushing those up in the definite win bucket.
Thinking of game outcome possibilities from a certain late inning point on can make us forget that the bigger chunk of wins and losses occurs before the 8th and 9th innings. Go up 4 by the time the 9th rolls around and you've robbed your closer of a chance to have a game that needs saving. And that's a good thing. So is taking a game that would have been tied and turning in to a save chance. But what happens when you get to those save chances? Well, it's a relatively small number of innings among teams that are relatively closely matched. Randomness is going to dominate.
Sure all gains are marginal. But if you want to know which teams from year N improved their W-L record in year N+1, you won't find it in their record in 1 run games. You will find it in their overall run distribution (or average run spread).