# So, which pitcher has the best pickoff move?

April 16, 2007 1 Comment

I suppose that this is technically the fourth part of my “Throwing to First” series. I was actually going to hold off on writing this one, but Tango Tiger suggested that I write it. I must admit, I’m not a fan of “best/worst lists”, but it is something of a natural progression, given all that we’ve discovered about the pickoff throw to first to ask who does it well and who does it poorly.

First a few points about pitcher characteristics. Handedness of the pitcher does not predict whether or he will make a throw to first (in fact, the percentages were identical!), or when he does, how many throws he will make. The chi-square and t-test were non-significant, respectively. However, when a pitcher does throw over, handedness counts in preventing stolen bases. Runners attempting to steal when a right-handed pitcher were successful about 77% of the time when he didn’t throw over and 69% of the time when he did. For lefties, runners were 75% successful when there was no throw and 53% when there was. A righty’s throw is only worth 8 percentage points on average, but a lefty’s is worth 22. Perhaps the fact that a lefty is constantly looking at the runner at first has something to do with this particular finding. When the pitcher isn’t paying attention, pitcher handedness makes no difference. If he is, lefties are generally better. In statistics, we call this a moderation effect.

Through my studies of the subject, it looks like the throw to first doesn’t deter runners from running, but it does slow them down, making a successful steal more difficult. A good throw to first should have three characteristics: It should reduce the runner’s chance of making it to second, potentially pick him off first, and not end up down the right field line. So, who does these three things the best in baseball? How do we measure such things?

First off, let’s look at slowing runners down. Caught stealing’s are difficult to assign fielding credit for, principally because they are the product of a pitcher, catcher, and whomever receives the throw at second base. A pitcher who has a lot of runners caught against him may simply have the benefit of having a strong-armed catcher on the other end of his pitches. (Think Pudge Rodriguez in the mid-90s.) But if we compare a pitcher against himself, then we can hold the effects of the catcher behind the plate roughly constant. If a pitcher is particularly good at holding the runners on, the spread between his stolen base success rate against when he doesn’t throw over and when he does should be comparatively large. In this case, the catchers behind the plate remain relatively constant, and the only thing that changes is whether or not a throw is made. The league-wide separation rate in the two scores is 11.4%. So, a pitcher who has a separation greater than this is doing better than average.

The average run value of changing a stolen base to a caught stealing is .608 runs (.169 for the SB, .439 for the CS). So, changing 10% of those stolen bases to caught stealings drops the run expectancy of the situation (ceteris paribus) by about .06 runs. If a pitcher has a separation of 20%, then the figure jumps to .12 runs. I calculated how much each pitcher reduced his stolen base success against rate when he threw to first vs. when he did and subtracted the league average of 11.4%. I then weighted his percentage above or below average by how many times a runner attempted to run after he threw over. To clarify: let’s say that runners stole successfully 71.4% of the time when no throw was made and 50% when he did throw. This is a separation of 21.4%, which is 10% better than the league average. If ten runners attempted to run on this pitcher after he made a throw, then one extra runner (over the league average) was likely converted into a caught stealing rather than a stolen base. The additional runners converted into CS over (or under) the league average was converted to a run value by multiplying by .608.

For pickoffs, I used a similar method. There were 276 runners picked off in 2006, about 8/10ths of a percent of all runners), and this reduced run expectancy by about 126.22 runs over the course of the year. So the average pickoff is worth .457 runs. Given how many times the pitcher found himself in the situation, I calculated how many pickoffs could be expected if he were a league average pitcher (e.g., I would expect a pitcher who had been in 100 such situations to have .8 pickoffs). I compared this to actual pickoff counts and assigned a run value based on the number of pickoffs above expected.

Finally, I calculated how many errors were made by the pitcher (I excluded errors charged to the first baseman), and found that they occured among 7/10ths of a percent of all runners on first (with second base open) and that the errors each cost an average of .258 runs. I calculated how many errors a league average pitcher would be expected to make (given the number of times the pitcher found himself in that situation) from actual error on pickoff throw totals and assigned a runs lost value to each pitcher.

The total runs saved over (or under) the league average was simply a sum of these three terms. Some pitchers could not have their statistics calculated, most often because no one bothered to run on them in both a situation where they had made a throw and a situation where they hadn’t. Most of the time these were relief pitchers.

And the best pickoff moves in 2006, measured by runs saved above average belonged to:

- Mark Buehrle (L) 8.11 runs saved above average
- Zach Duke (L) 6.13
- John Lester (L) 5.77
- Joe Beimel (L) 5.64
- Wandy Rodriguez (L) 5.60
- Justin Verlander (R) 4.39
- Jason Shiell (R) 4.39
- Jamie Moyer (L) 4.30
- Claudio Vargas (R) 3.72
- Scott Olsen (L) 3.63

On the other side, the worst pickoff pitchers of 2006 were

- Chen-Ming Wang (R) -4.10 runs saved
- John Smoltz (R) -3.65
- Chris Young (R) -2.50
- Roger Clemens (R) -2.32
- Greg Aquino (R) -2.28
- Roy Halladay (R) -2.26
- Juan Cruz (R) -2.25
- Runelvys Hernandez (R) -2.23
- Chad Billingsley (R) -2.12
- Jake Peavy (R) -2.07

This is a first stab at quantifying things. The fact that most (all?) of the pitchers on this list are starters did not escape me. Starters throw more innings and have more chances to save (and or allow) more runs. I briefly considered normalizing these figures based on either innings pitched or number of situations faced with a runner on first (the latter being preferred as the former would not control for the possibility that a pitcher simply allows a lot of guys to get to first base). However, I left the raw run totals as they were to show how valuable a guy like Mark Buehrle and his move to first actually are to a team. Just by his move to first, Buehrle saved the White Sox 8 runs. Add 2 runs resulting from confusion in how to spell his name and that’s a win right there for doing nothing other than keeping the runner at first close.

I’m also a little concerned because some of the stolen base success rates were based on pretty small sample sizes, although this was probably washed out somewhat by weighting based on how many runners actually ran against the pitcher in the “throwing” condition. Perhaps this is a stat more properly calculated across a few years worth of data. I’d be happy to hear critiques on this one or if anyone else knows of any other studies that have looked at the value of pickoff moves. If someone wanted to pick up this baton and run with it, you might check to see how stable these figures are across time. In any case, I present a work in progress. (And no, there won’t be a Part V.)

For those interested, the best pickoff moves, normalized by number of situations in which the pitcher faced a runner on first and second base open. Figures can be interpreted as how much a pitcher reduced run expectancy in that situation through his throws to first,as compared to the league average. (Minimum 25 such situations)

1) Joe Beimel .094 runs

2) Kiki Calero .083 runs

3) John Lester .076 runs

4) George Sherrill .070 runs

5) Mike Myers .059 runs

6) Chris Michalak .058 runs

7) Hong-Chih Kuo .057 runs

8) Mark Buehrle .051 runs

9) Wandy Rodriguez .050 runs

10) Jason Shiell .047 runs

A few relievers pop their heads in, although again, I worry about small sample sizes with them. Considering that the run expectancy from a runner on 1st with one out is somewhere in the neighborhood of .92, Beimel takes about 10% of that away.

On the other end of things, #10 worst move in the league belongs to Mariano Rivera (-.028 runs). Now, actually getting to first base with Mariano is another matter…