Projecting Jose De Leon

At long last, the Dodgers found a solution to their hole at second base, acquiring second baseman Logan Forsythe from the Rays on Monday night in exchange for top pitching prospect Jose De Leon. This came after months of rumors around a trade involving De Leon and Brian Dozier. The Dodgers had a surplus of starting pitchers and an opening at second, so it was only a matter of time before they dealt the unproven De Leon.

De Leon’s first crack at the big leagues — a four-start cameo in September — didn’t go quite as well as many had hoped. But he breezed through the minors over the last two years. He broke out in a big way in 2015, striking out an absurd 35% of opposing hitters between High-A and Double-A while walking just 8%. That performance made him a consensus top-30 prospect the following winter.

De Leon battled injuries in the first half of 2016, but began dominating again once he returned to the field. In 16 starts at the Triple-A level, he once again posted a strikeout rate well over 30%, along with solid walk and home-run numbers. De Leon proved himself at the highest level of the minors at the tender age of 23. Pitchers who meet that standard often go on to have success in the majors, especially when they miss bats as prolifically as De Leon did.

De leon grades out exceptionally well by my KATOH system. It projects him for 8.1 WAR over his first six seasons by the traditional method (KATOH) and also 10.1 WAR by the method that integrates Baseball America’s rankings (KATOH+). He’s the 13th-highest-ranking prospect by KATOH+ and the third-highest-ranking pitcher.

To help you visualize what his KATOH projection entails, here is a probability density function showing KATOH+’s projected distribution of outcomes for De leon’s first six seasons in the major leagues.

To put some faces to De Leon’s statistical profile, let’s generate some statistical comps for the newest member of the Dodger’s rotation. I calculated a Mahalanobis distance between De Leon’s performance this year and every Triple-A season since 1991 in which a pitcher recorded at least 350 batters faced. In the table below, you’ll find the 10 most similar seasons, ranked from most to least similar. The WAR totals refer to each player’s first six seasons in the major leagues. A lower “Mah Dist” reading indicates a closer comp.

Please note that the Mahalanobis analysis is separate from KATOH. KATOH relies on macro-level trends, rather than comps. The fates of a few statistically similar players shouldn’t be used to draw sweeping conclusions about a prospect’s future. For this reason, I recommend using a player’s KATOH forecast to assess his future potential. The comps give us some interesting names that sometimes feel spot-on, but they’re mostly just there for fun.

Jose De Leon’s Mahalanobis Comps
Rank Name Mah Dist KATOH+ Proj. WAR Actual WAR
1 Rich Harden 2.16 6.6 15.1
2 Anthony Reyes 2.44 6.5 1.2
3 Hayden Penn 2.56 8.0 0.0
4 Ruben Quevedo 3.31 6.0 0.0
5 Homer Bailey 3.31 5.5 13.1
6 Wade Miller 3.37 5.2 13.2
7 Matt Cain 3.46 11.2 24.0
8 Carlos Carrasco 3.51 6.5 9.6
9 Dan Reichert 3.56 8.5 2.2
10 Juan Pena 3.73 10.5 0.7

For all his minor-league dominance, De Leon does have his flaws. Consider Eric Longenhagen’s comments from his write-up of the Dodgers system:

DeLeon’s drop-and-drive delivery sucks plane out of the fastball, and he’s fly-ball prone when he’s working up in the zone… His low-80s slider is an average offering and only consistently effective when he’s locating it just off the plate to his glove side. De Leon didn’t do this in his brief major-league stint.

Those detractors help explain why he got a little roughed up last September, and will be worth monitoring next year. But all in all, De Leon has been one of the top performers in the minors the past couple of years and is clearly ready for the next challenge. In De leon, the Rays now have an arm who can strengthen their rotation right away, and could provide significant value in the longer run.





Chris works in economic development by day, but spends most of his nights thinking about baseball. He writes for Pinstripe Pundits, FanGraphs and The Hardball Times. He's also on the twitter machine: @_chris_mitchell None of the views expressed in his articles reflect those of his daytime employer.

6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Jimbomember
5 years ago

Hey Chris, maybe a stupid question, but what causes the KATOH output to not be more bell shaped? As in, De Leon has one outcome that’s most likely, and outcomes become less likely as you move away from that. I’m reacting in particular to the dip at 13-16 WAR and subsequent rise at 16-20 WAR.

piddy
5 years ago
Reply to  Jimbo

KATOH uses multiclass probit regression. That means it’s not actually estimating a probability distribution over the WAR results, but rather probabilities of falling into certain bins (that’s also why the plot of the “probability density function” isn’t a density function at all). That means you can’t really enforce a sensible assumption like unimodality.

someone
5 years ago
Reply to  piddy

Enforcing unimodality might be difficult, but you can still apply more structure. For example, the ordered probit:

https://en.wikipedia.org/wiki/Ordered_probit