Maintaining Velocity

A few weeks ago I took a look at pitchers that saw significant changes in their average fastball velocity from '08 to their first few starts of '09. At the end of that post, I wondered how sustainable those velocity increases/decreases were going forward, and how much of it was just noise.

As it turns out, those changes were quite sustainable. The regression below looked at all the guys in the original sample who have faced at least 60 batters since my initial analysis. It shows how each pitcher's fastball velo in weeks 4-6 of '09 is best predicted by a combination of their '08 FBV, and their velocity from the first three weeks of this season.

This was a very encouraging sign (unless you're Daniel Cabrera, I guess). After just three weeks, a pitcher's 2009 velocity is nearly three times as relevant as his '08 velocity. This was not terribly shocking, since velocity stabilizes so quickly, but the magnitude is impressive.

Another thing we can do is look at changes in velocity in conjunction with other improvements, like strikeout rate. Is a pitcher striking out more guys and experiencing a jump in velo more likely to maintain that K/9 increase than a guy throwing the same speed as last year? As it turns out, yes:

The first row tells us that for all pitchers in the sample, after three weeks we should use about 70% '08 K/9, and 30% '09 K/9. When we break it up into guys who have seen big changes in velo and guys who haven't, things become a little more clear. For the guys who have experienced at least a 1% increase or decrease in velocity, the ratio of '08 to '09 drops to 60/40. That makes sense, as we've got more info suggesting that something is inherently different with them.

But if the pitcher is throwing approximately the same speed as they were in '08 (+/- 1%), it seems more likely that the jump in K rate was just noise. For them, the ratio of how much we should weight '08 and '09 jumps up to 83/17, and the '09 coefficient actually isn't even close to being statistically significant, with a p-value of 0.41.  And it's not like we're talking about ERA here; people legitimately get excited, and for good reason, when a guy bumps up his K rate early in the year.  But it seems like that excitement is not always warranted.

That two pieces of info tell us more than one isn't surprising of course, but it's always nice to quantify something that intuitively makes sense. Tomorrow I'll take a look at the 10 guys we focused on a few weeks ago, and see how telling their changes in velocity ended up being.

All data from the great FanGraphs.