/cdn.vox-cdn.com/uploads/chorus_image/image/48841553/usa-today-8836687.0.jpg)
Last week, we looked at the Steamer and ZiPS projections for the Rays. Very soon we'll use the projections to build a batting order, but I thought it was worth taking a post to talk just about splits. Overall projections are great, but if want to know the outcome of a specific at bat, we can do better. We should, at the very least, consider handedness splits.
The basic idea is that almost all batters hit opposite-handed pitchers better than they hit same-handed pitchers, but the degree to which players hit differently against pitchers of different handedness varies, and is a part of each batter's profile. We can't, however, just look at a players career splits and project that pattern to continue going forward. That's not something we'd do with overall production, so we won't do it with splits, either. We have to regress.
Regular readers of the site will already know how this works, so bear with me for a minute. Jason Hanselman (of The Process Report) and I jointly made a tool that does the regression quickly for us, but the research is not ours. The concept was first flushed out in The Book, but the numbers we currently use are from Bojan Koprivica's work on The Hardball Times. Sometimes Jason and my numbers will differ, but that's usually due to using a slightly different overall projection, or a slightly different way of handling park effects. Here are some basics to remember:
- The average right-handed hitter split is .023 points of wOBA.
- The average left-handed hitter split is wider at .035 wOBA.
- Generally, you should regress against 1500 PAs against left-handed batters(there's a more complex way to think of this, but no reason to go into it -- if you're interested, read Koprivica's work).
- Below is a chart of career plate-appearances for all of the Rays players. Not a single one of them has yet reached 1500 PAs against left-handed pitchers. That's the threshold where we can start thinking that the observed value is 50% signal. This is why regression is crucial.
Player | PAs | PAs vs. RHP |
PAs vs. LHP |
James Loney | 5120 | 3799 | 1321 |
Evan Longoria | 4789 | 3468 | 1321 |
Logan Morrison | 2355 | 1698 | 657 |
Desmond Jennings | 2126 | 1524 | 602 |
Logan Forsythe | 1713 | 1115 | 598 |
Steve Pearce | 1555 | 961 | 594 |
Brad Miller | 1243 | 908 | 335 |
Hank Conger | 997 | 811 | 186 |
Rene Rivera | 992 | 699 | 293 |
Corey Dickerson | 925 | 728 | 197 |
Kevin Kiermaier | 899 | 670 | 229 |
Brandon Guyer | 729 | 339 | 390 |
Nick Franklin | 611 | 426 | 185 |
Steven Souza | 452 | 325 | 127 |
Tim Beckham | 231 | 112 | 119 |
Curt Casali | 197 | 144 | 53 |
Mikie Mahtook | 115 | 38 | 77 |
Richie Shaffer | 88 | 25 | 63 |
Luke Maile | 35 | 14 | 21 |
Now let's get to the actual numbes.
While PECOTA's projections came out today, I'm just using the aggregate of Steamer and ZiPS. That's because they're fully public, and they both express their overall projection on the same scale (wOBA, as opposed to TAV). You can and should look over the PECOTA projections to see where they disagree, but on the whole they're pretty similar. These are projected for a season with half the games in Tropicana Field (a pitchers' park)
First off, here are the splits with pretty colors.
And here they are with sortable tables. Take a second to sort by the projections for each handedness.
Player | "Handedness" | "Proj. wOBA" | "Proj. wOBA vs. RHP" | "Proj. wOBA vs. LHP" |
---|---|---|---|---|
Corey Dickerson | L | 0.326 | 0.335 | 0.294 |
Evan Longoria | R | 0.323 | 0.315 | 0.343 |
Steve Pearce | R | 0.323 | 0.314 | 0.338 |
Steven Souza | R | 0.323 | 0.316 | 0.339 |
Logan Forsythe | R | 0.318 | 0.306 | 0.339 |
Brandon Guyer | R | 0.316 | 0.302 | 0.327 |
Brad Miller | L | 0.314 | 0.324 | 0.284 |
Kevin Kiermaier | L | 0.314 | 0.324 | 0.284 |
Logan Morrison | L | 0.312 | 0.320 | 0.290 |
Desmond Jennings | R | 0.312 | 0.304 | 0.330 |
James Loney | L | 0.303 | 0.312 | 0.274 |
Nick Franklin | S | 0.299 | 0.305 | 0.284 |
Richie Shaffer | R | 0.292 | 0.282 | 0.296 |
Curt Casali | R | 0.290 | 0.285 | 0.303 |
Hank Conger | S | 0.289 | 0.293 | 0.273 |
Mikie Mahtook | R | 0.288 | 0.276 | 0.294 |
Tim Beckham | R | 0.276 | 0.268 | 0.284 |
Luke Maile | R | 0.263 | 0.252 | 0.270 |
Rene Rivera | R | 0.260 | 0.255 | 0.272 |
Some observations:
- The Rays really struggled against right-handed pitching last season, and the front office tried to address that in the offseason. Three of the four top projections against righties are new acquisitions.
- The second-highest projection against RHP is Kevin Kermaier, which may surprise some folks. It surprised me. He was also a player that ZiPS and Steamer disagreed on, with Steamer giving him a .307 wOBA overall and ZiPS a .320 wOBA. If you're wondering, PECOTA falls on the optimistic side, with a .263 TAV. That would be a good outcome for the Rays.
- Evan Longoria shouldn't be expected to carry the team against righties, and maybe this season he finally won't need to.
- Steve Pearce, a righty, has a higher projection against right-handed pitching than does James Loney, a lefty. If that's correct, that makes it pretty hard to find a spot for Loney on this roster, as Pearce is probably best-suited for first-base defensively.
- Sorting by projection against lefties tells a different story about the other Ray often discussed in trade speculation. Desmond Jennings projects to be above average, the fifth-best hitter against LHP on the team, and the second-best outfielder against them. Maybe he should be traded, but either on this team or on another, there is value there.
- I think a lot of people expect Tim Beckham to beat his overall projection, and to be a decent player vs. lefties. Right now though, the projections have him on the same level as Brad Miller and Nick Franklin, both of whom are likely superior plays against righties.
- Catcher platoons aren't the most exciting offensive duos, but at least Conger and Casali match up well in that regard.
The next article in this series will take this data and attempt to build some possible Rays batting orders.