clock menu more-arrow no yes mobile

Filed under:

Examining possible ranges of outcomes: Starting pitchers

Looking at the best and worst case scenarios for the rotation.

Boston Red Sox Spring Training Workout Photo by Billie Weiss/Boston Red Sox/Getty Images

We are now into what should have been the third full week of the major-league season, which means we are getting pretty deep into this whole baseball-less thing. That also means that, even if baseball does come back, it is going to be very different than anything we’ve ever seen before and thus we can pretty much throw many or most of our preconceived notions about things out the window. I guess what I’m trying to say is that I’m just not sure how useful projections will be this year in the event baseball is played.

With that out of the way, I’m going to spend the rest of this article talk about projections, because what else do we have?! There are a lot of different projection systems out there these days, which is a good thing because no one system is perfect but the more data points you can throw into a sample the more useful the information you do have should be. Baseball Prospectus’s PECOTA system is one of the most famous systems and has been around forever at this point, and again while it’s not perfect it is one of my favorites.

Specifically, I love that PECOTA doesn’t just publish the 50 percent projection. That is still their main output, but BP will also publish different percentile projections for every hitter, essentially creating a ceiling and a floor for a player. They go mostly by tens of percentage points, but also include the one percent range and 99 percent range projections. Essentially, using their model they simulate the season tens of thousands of times and these are the performances that happened 90 percent of the time, 10 percent of the time, et cetera.

I’m going to take a look at those ranges for the players on the Red Sox roster (click here for BP’s depth chart and links to individual player projections) today and see what the more relatively realistic possible range of outcomes are for everyone, looking at the 10th and 90th percentile outcomes. Today I’ll look at the starters, with the relievers coming at a later date.

Eduardo Rodriguez

10th Percentile: 0.3 WARP

90th Percentile: 2.1 WARP

When the de facto ace of the staff is only projected to be a two-win pitcher by the rosiest of outcomes, that signals an issue for your rotation. That being said, I should mention that part of this comes down to innings. Rodriguez threw over 200 innings for the first time in his career last season, which is great, but no projection system will project that moving forward after just one year. That’s just not how models work. So, this 2.1 WARP comes over 157 innings. If we project out to 200 innings, it bumps up to 2.7, which isn’t super impressive in its own right but is better. Baseball Prospectus’ pitching WARP is based on DRA, which in mind is the best pitching metric insofar as its ever right to only use one. They predict a 90th percentile performance to be ten percent better than league average by this metric — for context, he finished with a 91 DRA- last year — and a 10th percentile outcome being 11 percent worse than average. The big difference comes down to walks, with his per-nine rate jumping from 3.0 to 4.2.

Nathan Eovaldi

10th Percentile: 0.2 WARP

90th Percentile: 1.8 WARP

Again, this comes partially down to innings. In fact, on a per-start basis Eovaldi’s best-case actually comes out slightly better than that of Rodriguez, with his 90th percentile DRA- coming in one point better than the lefty’s at 89. This projection has Eovaldi at only 134 innings, which would actually be his highest regular season total since 2015. If we project out to 200 innings — not that I expect that to be realistic, just to make a point — it would be pretty much exactly the same was Rodriguez’s, edging him out by a hundredth of a win. In fact, the similarities between the two don’t stop there, with their margin of difference in both strikeouts and walks between the two percentiles being pretty close to equal as well.

Martín Pérez

10th Percentile: -0.7 WARP

90th Percentile: 0.9 WARP

Yikes! Not sure what else to say here. Well, yes I do. This could be seen as a flaw in projection systems, to be sure. Now, I am not exactly high on Pérez and not a huge buyer into the changes he made last year transferring into success in 2020. That said, there were tangible changes in the repertoire that led to a massive amount of weak contact, which is the type of data that can be hard to put into a projection model under the proper context. DRA has never liked Pérez — he’s been at least 30 percent worse than league-average by DRA in each of the last five years — which explains the 90th percentile projection here being a 103 DRA-. There’s always reasons to talk yourself out of a projection, but the case might be easiest here if you are buying into the changes he made last year despite the results not following them. If you think that was just bad luck mixed with a need to refine a few things in a second year with the new arsenal, Pérez is the guy to look at as someone who can beat their 90th percentile projection.

Collin McHugh

10th Percentile: 1.0 WARP

90th Percentile: 2.7 WARP

According to PECOTA, the best pitching in the Red Sox rotation is actually McHugh. Now, I have a little trouble with this one because of how much of his recent data is out of the bullpen. In fact, he is projected to make 19 starts and 29 relief appearances in this projection, though it totals out to 151 innings. As a starter, he hasn’t been quite as effective over the last few seasons as he has been out of the bullpen, and that’s without even mentioning that he’s coming off an injury that he is still in the midst of rehabbing. None of that is to say McHugh shouldn’t start, because, well, *gestures at the Red Sox rotation depth* but rather that I don’t see him being safer than Rodriguez or maybe even Eovaldi for that matter. They have his tenth percentile outcome being league-average. If he’s a pure starter, I certainly don’t buy that.

Ryan Weber

10th Percentile: -1.6 WARP

90th Percentile: -0.7 WARP

I really have nothing to say here. I would disagree with the ceiling here in that I think there exists a best-case scenario where Weber is replacement level. Even PECOTA’s 99th percentile projection has him slightly below replacement level, and I should mention this is over 72 innings! If you extend to 200, his 10th percentile projection is -4.4, which is basically unfathomable. Of course, if he was that bad they wouldn’t give him 200 innings. Anyway, even if I disagree with the extent to which this projection is terrible, it speaks to the fact that this roster has major issues at the back of its rotation.


10th Percentile: -0.8 WARP

90th Percentile: 6.8 WARP