Dave Cameron over at Fangraphs on Josh Beckett's amazing resurgence:
Last year, Beckett had an xFIP of 3.86, 8% below league average. This year, Beckett is posting a 3.69 xFIP, 8% below the league average. In fact, his K/BB ratio is almost exactly identical (2.58 last year, 2.63 this year) to what it was a year ago. His ERA has been slashed by over four runs thanks to huge reductions in two factors that are counted in xFIP – BABIP and HR/FB. …
This year, Beckett has the lowest BABIP (.217) of any starter in baseball, and his 3.9% HR/FB rate is the fifth lowest of any qualified starter. He’s regressed right past the mean, and now his performance in 2011 is as unsustainble as his 2010 performance was, just in the other direction this time. Just like Beckett was a great pick to improve upon his struggles last year, he’s a good bet to regress in the second half of this year. …
Josh Beckett was never terrible, and he’s not amazing now. More than anything else, he’s an example of why ERA isn’t a good tool for projecting future pitching performances.
This is true to an extent, but as Marc Normandin points out in the comments, Beckett was so hittable in part because his back prevented him from properly throwing his curveball, not only because he was unlucky with balls in play and home runs.
Aside from that point, there's a conflation in the final paragraph of "projectability" with "amazingness/terribleness."
In the end, pitching is about results, and while I agree ERA is not useful in isolation to predict future performance, it is a decent measure of how well a pitcher handled his primary job of preventing runs. In the end, a 5.78 ERA is terrible and a 1.86 ERA is amazing.
This is probably my No. 1 beef with Fangraphs — perhaps the only one beyond their hideous green-themed website, truth be told. Their WAR figures for pitchers are based on FIP, rendering useless a statistic purporting to assess a player's value because it ignores how valuable a pitcher actually has been.
It's not that I disagree with discrediting or devaluing a lot of the traditional statistics, like batting average, wins and RBI. But that's because those categories incorrectly or incompletely assess current performances, regardless of projectability. But this is a whole new level: The dismissal of actual results in assessing a player's value. Josh Beckett was terrible in 2010. He is amazing now. Those are completely different assessments from whether he will be either of those things in the future.