Sabermetrics predicted nothing for the Red Sox. Sabermetrics suggest that the easiest way to improve your team is by improving the defense and the pitching. If the players don't perform, that says nothing about sabermetrics.
Nate Silver, a performance analyst and the creator of Pecota, projected the Sox to win based on his performance projections for the players. I'm not going to go into how he goes about it as it's way too complex to explain in few words. Thus far, the players are not performing up to expectations and, of course, Pecota can't project injuries. And, of course, no performance analysis system is going to project performance over a three week stretch. In fact, both Lester and Beckett are performing exactly as they did last April. Lester spent the final four and a half months as the best pitcher in the AL and Beckett improved enough to wind up with a decent, if slightly sub-par season for him.
Right now, they're only four games behind the Yankees, despite the fact that the Yankees have four starters performing at a level they can't possibly maintain. One thing that has the Sox where they are is...luck. One piece of sabermetric analysis is Pythagorean standings, which are based on teams' run differential. Both the Rays and Yankees have the records their diffential suggests they should have. The Sox have won one more game than they should have. The Sox have won 5 straight one run games. Luck has been on their side.
And, by the way, knowledgeable Sox fans aren't worried. You don't worry in April.
