THE BOOK--Playing The Percentages In Baseball

The main thing I got from that was that we need a bigger sample size. But it is an interesting thought process nonetheless.

Pitch Runs above average
Type Per 100 pitches
Fastball 2.07
Slider -0.59
Curveball -0.87
Change-up -1.56

Please, please, please, when you give us numbers like that, can you also give us league averages?  How in the world are we supposed to know whether this batter is good or bad on the various pitches if we don’t know what the average batter does? “Good” and “bad” are always relative to league average. That would be like giving us this piece of data:

Batter A

120 mph fastball: -3.56 runs

Last year, the average pitch speed on caught-stealings was 87.1 mph; on successful stealing attempts it was 86.6. The success percentage was 75 on fastballs, 80 on sliders, 82 on change ups and 84 on curveballs.

Great stuff!

This year Tampa Bay has been opening its offensive first innings with Upton, then Crawford, Longoria and possibly Pena and Burrell. Thus, the steal threat and the fastball crusher have been penciled back-to-back on the scorecard many times.

I have no data handy, but Upton has been batting at the bottom of he order for quite some time now.

Many analyses have been done on lineup optimization. While we have many tools for building an optimal batting order given nine hitters, this is an attempt at assessing how much a good runner on first can affect the production of a batter. The bottom line is… very little.

I like the attempt!  Now, with the difference between the SB success rate of the fastball and curve ball being 9%, and the fact that on any given pitch, the chances of the runner going is around 10% or so (maybe less - I am guessing), I wouldn’t think that the runner should influence the pitch selection very much at all.  It definitely changes everything (because your pitch selection in the first place is dictated by the run value of each pitch), but not by much.

Basically if on any given pitch, the runner is going to be running 10% of the time, and even then, it only matters when the batter does not swing, we have a difference in run value of around .005 runs for a SB attempt (between a fastball and a curve ball). Not enough to change your approach very much.

Also, Max, please give us some more sample sizes so we can put these numbers in context!

Tampa batting lineups, game-by-game:
http://www.baseball-reference.com/teams/TBR/2009-batting-orders.shtml

It was Upton mostly, then went to Bartlett mostly.

I love this page, as well as its sister fielding lineups:
http://www.baseball-reference.com/teams/TBR/2009-lineups.shtml

Estimating the averages from FanGraphs’ 3-year leaderboard, I get something like the following for league averages per 100 pitches:

FB:  0.19
SL:  -0.55
CT:  -0.48
CB:  -0.13
CH:  -0.08
SF:  -0.21
KN:  0.33

Those could be off, though, because I estimated number of pitches from total run values and values per 100 pitches for each player, and I suspect the rounding could throw off some of the estimates, especially when a player was right at average.

I agree that that is important for seeing if Longoria is actually a good fastball hitter or not, but I don’t think it’s that big a deal that it’s not included (especially since FanGraph’s doesn’t give the averages).  For one, I would expect the averages would be pretty close to 0 for each pitch just because of the game theory:  as a whole, you would expect pitchers balance the likelihood of throwing each pitch so that they are all around the same value.  If a pitch is easier to hit, they should throw it little enough that the hitter is thrown off a little more by seeing it and it ends up around the same value per pitch.  In practice, the averages probably aren’t exactly zero, but are probably close enough to it that a split as big as Longoria’s last year is easily bigger than average.

Also, I’m not sure it’s even that important for the point of the study (beyond just how you would use it for regression if you took it that far).  It’s important to see if Longoria was really a fastball hitter moreso than normal last year, but that’s not really an important point to the study.  The extra value of seeing a few more fastballs just depends on the relative value of each pitch for him, not whether his splits are league average or not (regression aside).  It would be nice to see, but I don’t think it’s really important to the study beyond that one trivial point.