I'm curious - how do you come up with this confidence interval? Is it just some kind of t-test using 1's and 0's for the first 30 games?

The Yankee turnaround really wasn't so amazing. At their 11-19 low, the team's run diff was 152-173. Not so hot but significantly better than the won-lost record. Writing the Yanks off was more wishful thinking than unbiased analysis.

The Yankees were scoring runs: 5.1 a game even at the low point. So if they got halfway decent pitching, they could always put up a lot of wins. In the second half NY's starting pitching improved to so-so from awful. And guess what. The team put up a lot of wins.

Cano had little to do with it, by the way. His .779 OPS is an improvement over Womack and whoever. But he's hardly a mainstay of the Yankees' offense (eighth among the regulars in OPS). The Yankees' offense was just fine before he arrived and has continued to be just fine.

But Wang, Small and Chacon did have a lot to do with the improvement in NY's rotation. And that was the critical factor - along with less bad luck compared to their run diff - which brought the Yankees around.

Unrelated note: MLB will set an attendance record this week, barring rainouts of the entire remaining schedule. They need about a million more tickets sold to get the record, which should fall Thursday.

The record is entirely due to the move from Montreal to Washington. Without those extra two million tickets, it looks like MLB would actually have posted an attendance drop this year of several hundred thousand. Fouteen of the other twenty-nine teams are down in per-game attendance this year.

I'm not so sure you can write off Cano so easily. You underestimate the value of replacing Womack with an adequate performer.

The difference is insignificant. In games where Cano didn't play, the Yankees averaged 5.43 runs per game. In games when Cano did play (almost always as a starter) the Yankees averaged 5.45 runs per game.

The Yankees scored pretty much the same with Cano or without him. Frankly, I wouldn't expect much difference from a .294/.318/.462 hitter. The OBP is subpar even if the SLG looks pretty good. The overall Yankee team stats are .276/.355/.450. Cano pretty much matches his team all-round. He doesn't hurt them much, he doesn't help them much.

By contrast, Chacon, Small and Wang have put up ERAs of 2.89, 3.25 and 4.02 compared to the Yankees' overall 4.44. Those are pretty significant differences, and they really tell why the Yankees starting pitching improved so much in the second half.

Don't forget the acquisition of Jay Payton at the All-Star break for the A's. They had already turned their season around, but Payton's play in the second half of the season helped them stay hot and has helped them stay in the race in the last month as their rookies have struggled and others got injured.

Don't forget about the fact that Oakland's starting pitchers really came forward - Zito got hot and Blanton's K rate went way up. I think that this account for a lot of their success, although the contribution of Payton and the rookies (Dan Johnson, Swisher) are pretty significant as well.

Using the binomial distribution the probability of a .587 team going 11-19 at the start of the season is .0123, not .0044. Unlikely yes, but not really remarkable. Note that the probability of a .587 team going 11-19 at any point in the season is much greater, since the bad stretch could begin with any game from #1 through #132.

Of course, the larger issue is, were the Yankees a .587 team at the start of the season? That question can't be answered statistically. All we can say is that their win probability is constantly changing, not only because of changes to their lineup but because of varying strength of schedule.