THE BOOK--Playing The Percentages In Baseball

"I can only presume that in addition to the speed gain they have as relievers (say 2-3 mph), they have an equal (or better) gain in movement of pitches, plus of course the advantage of not facing the same batter multiple times.”

To test these, maybe you could:

“the speed gain they have as relievers (say 2-3 mph)” - look at only the AB as starters where they did have the same average fastball velocity as when they were relievers?

“the advantage of not facing the same batter multiple times” - compare the first time through the lineup as a starter vs. all relief appearances?

“hey have an equal (or better) gain in movement of pitches” - should be easy to confirm if this is true or not…

Also perhaps the k-rate is decreasing as the game goes on for starters, as they tire.

“the advantage of not facing the same batter multiple times” - compare the first time through the lineup as a starter vs. all relief appearances?

Not good enough, since the starter paces himself from the beginning.

Also perhaps the k-rate is decreasing as the game goes on for starters, as they tire.

I have it in The Book.  The K rate did drop somewhat each time through the order.

Another advantage most relievers have is that they often only need to throw their 2 or 3 best pitches, and not resort to their 4th or 5th best pitches (or at least not nearly as often) just to show the batter something different.

FYI: There’s an article at THT today with some evidence that additional velocity lowers BABIP:
http://www.hardballtimes.com/main/article/pitch-speed-and-balls-in-play/

Guy, you’re going to slap your head when you see the link at the top of this thread…

He groups the pitchers by average speed of all pitches (not just fastball), and then finds that the fastest group gets more grounders and has a lower BABIP. 

Wouldn’t that just be an artifact of sinker pitchers throwing mostly sinkers (the pitch mix results in a high average speed), who by definition need a low BABIP to stick in MLB?

Good comments so far here and on Ballhype.  Someone on ballhype pointed out that high K pitchers will have fewer runners on base when a ball is hit and that BABIP is a lot lower with no one on base.  Excellent point, which makes me think that for these BABIP studies, it is often necessary to control for baserunners. Excellent point!

And as Zack says, there could be selective sampling biases in these results.  I don’t like the idea of Dan using speed of all pitches combined as he is getting a combination of pitch speed and pitch selection in his groupings.  So to call any relationships he finds that of “pitch speed” and X, is misleading (not intentionally of course), even though it is technically true.  A pitcher like Maddux (or any of the sinker ball pitchers, as Zack says) is going to have a much higher average speed than a flamethrower who also throws lots of offspeed pitches or who throws a very slow curve.  In fact, I really don’t think we can use a pitcher’s average speed of all his pitches as a proxy for how hard he throws.  I realize that Dan did this intentionally, but because of all the other variables and the potential for selective sampling effects, I don’t think the the conclusions are necessarily (that the harder one throws, the more that batted balls are hit weakly) accurate.

And of course, if that were true to any significant degree, then DIPS would not have been right.  Since I assume that there is a high year to year correlation for a pitcher’s pitch speeds, then if pitch speed were significantly related to BABIP, it would follow that there would be a significant y-t-y relationship for BABIP as well - which we know there isn’t.  So something must be wrong with that conclusion.

I would not trust the relationship between K rate and pitch speed for the same reason.  That is not even an interesting relationship, again, if we are using speed of all pitches.  I don’t think we want to mix speed of fastballs versus K rate with speed of offspeed pitches versus K rate.  They may have opposite relationships/correlations so we certainly don’t want to mix the two.

I didn’t pay attention that it was all pitches.  Then again, I should have realized that given that the average pitch speed was around 87.5.

***

The way I select “fastball speed” is to take the 25% fastest pitches for a pitcher.  I think that’s a fairer way to do it than to take all of Trevor Hoffman’s fastballs (as little as he throws them), and all of Mariano Rivera’s fastballs (cutter), and then take the average of each.

What if Mo were to intentionally throw his fastball slower.  I think it is more likely for him to do that than Trevor.

This is similar to taking the average distance of a HR.  What if a power hitter, in addition to his long 30 bombs, also hits 20 just-overs.  His “average” would be depressed by those 20.