THE BOOK--Playing The Percentages In Baseball

Tango,

Let’s check by age…

Among those under age 30 in 2006, the regression equation is (2007 fb vel) = (0.96)*(2006 fb vel) + 0.04*(lg-avg fb vel)

(where the lg-avg fb vel is evaluated only for those pitchers in the new sample: under 30 in 2006, at least 100 IP in both 2006 and 2007.)

So, the regression is actually greater, which is surprising, but you are regressing to a higher fb velocity (91 mph for the young pitchers as opposed to 90 mph for all the pitchers in the original sample).  The r2 is 0.94, slightly higher than what I reported for the original sample.

You’re right, there’s a lot to do here.  This is barely scratching the surface.  But this is the sort of building block we’ll need in order to expand the “roots” of the sabermetric hierarchy.

This would seem to imply that the fastball for the over30s are going to be regressed to AWAY from the population mean, which is frankly impossible.  Unless of course, those are the guys who get more injured, and therefore, would have had lower FB scores in the previous year.

***

Another way to get fastball speed, which I think I’ve mentioned at some point in the past, is to take the top 25% of all pitches with the highest speed.  After all, what if one guy throws 40% fastballs, and another throws 70% fastballs, but the second guy has alot of “slow” fastballs in there (he purposefully throws it 1 or 2 mph faster, so that he can get better location or impart a tiny bit more spin).

By focusing on the top 25% of all pitches thrown (fastball or not), we can get around this.  Plus, if someone is hurt, we can be pretty sure that the top 25% of all his pitches thrown will be the pitcher at his “best”.  If not top 25%, then top 10%.

After all, the standard deviation of the fastball speed must be around 1mph (if you throw 90, then 95% of the time, your fastball will be 88-92).  So, we don’t really need to have 1000 pitches from a guy to know how fast he throws.  Just take his top 100 fastest pitches, let’s say, and that’s enough.

(Wakefield will of course be an exception.)

This would seem to imply that the fastball for the over30s are going to be regressed to AWAY from the population mean, which is frankly impossible.  Unless of course, those are the guys who get more injured, and therefore, would have had lower FB scores in the previous year.

Huh.  You’re right.  Not sure how to explain that, aside from your selective sampling argument.

I asked this of someone else and he didn’t get back to me, so I’ll ask here. Can you look at fastballs from 2007 and see how fast those are being thrown so far in 2008, while paying closer attention to cold weather climates? I looked at a handful of pitchers on Fangraphs and noticed that all of them have decreased fastball velocity compared to last year (Haren, Oswalt, Sabathia, Zambrano, Matsuzaka, Vazquez, and others).... can you see just how much velocity the hard-throwing pitchers have lost? I have a feeling it’s as much as 1-2 mph on average.

Dan, The average fastball velocity is down just a tad this year: 90.268 vs. 89.963.  Of the pitchers who threw over 92 last year, there are only few who have increased their fastball speed:  McGowan, Papelbon, Accardo, Colome, Casilla, Kery Wood, Aardsma, Smoltz, Embree, and Ervin Santana.

All the rest are down and every pitcher I looked at had at least 50 fastballs recorded in both seasons. 

But if you look through just the first couple weeks of both seasons, the majority of players are within 1 mph of where they were last season.  (Same 50 minimum).

I might be able to break out pitchers by cold weather at somepoint, but I just don’t have the data right now.