clock menu more-arrow no yes

Filed under:

Orioles Opening Day win another example of starting pitchers outperforming projections

New, 9 comments

On Opening Day, Chris Tillman picked up where he and other Orioles starters left off, outperforming their FIP.

Kim Klement-USA TODAY Sports

It's easy to bash the projections (especially those for the Orioles) because the Orioles have a recent track record of flipping those projections on their head and burying them in the outfield. (See this article by CC's Bill Duck from yesterday that gets into the history on that). Of course, here we are again in 2015 with the Orioles once again projected to finish last in the division. But why? What is it in the numbers that keep causing this to happen?

Looking at the 2015 Orioles ZiPS projections on Fangraphs, the team is supposed to score approximately 690 runs and allow approximately 704 runs. Compare that to last year's 705 runs scored and 593 runs allowed. The offense doesn't change much, but its the defense (and really, the pitching) that's projected to regress.

The issue is that while actual wins and losses on the field are determined by actual runs scored and allowed (runs allowed being measured mostly by ERA), the ZiPS projections estimate ERA using FIP, which can vary a lot from a pitcher's actual ERA. For example, look at Chris Tillman's ERAs over the last three seasons:

Chris Tillman ERA

Year

ERA

2012

2.93

2013

3.71

2014

3.34

You might average those to project his ERA for the 2015 season and get about 3.33. That's pretty good. But if you go look at his 2015 projected ERA on Fangraphs, it's at 4.11, almost a full run higher. Why? Because his FIP for the last three years (4.25, 4.42, 4.01) have an average of 4.23. His projected FIP for 2015 is 4.40 and his ERA is projected to be 4.00, not  3.33. Assuming he pitches 200 innings, that's a difference of 15 runs just from one pitcher. I'm not saying averaging a pitcher's ERA is a great way to predict his future performance, but intuitively it might make more sense and explain how someone's perception of a player can differ so much from a projection, and how that individual projection can affect a team's overall projection.

This isn't unique to Tillman either:


2014 ERA

2014 FIP

2015 Proj. ERA

2015 Proj. FIP

Gausman

3.57

3.41

4.12

4.05

Jimenez

4.81

4.67

4.19

4.34

Gonzalez

3.23

4.89

4.40

4.92

Chen

3.54

3.89

4.10

4.10

Norris

3.65

4.22

4.21

4.33

You'll see that with the exception of Jimenez, everyone's FIP is higher than their ERA last year and the ZiPS projections use FIP in their projections. So again, those projections are assuming the Orioles are going to give up more runs then you might expect based on watching this team from last year. The people who build these projection models have their reasons for doing it this way, and I don't really want to get into that nor do I necessarily disagree.

You'll see for Tillman, Gonzalez, Chen, and Norris that their FIP is greater than their ERA. This implies a certain amount of luck (or very good defense) was in play and that the theory is there should be some regression toward the mean and the pitcher's performance will not continue to out-perform their FIP over the long term. But were the Orioles pitchers performances last year so lucky that it should imply a lot of regression this year? Here's the difference between each pitcher's ERA and FIP:


2014 ERA

2014 FIP

Difference

Gausman

3.57

3.41

0.16

Jimenez

4.81

4.67

0.14

Gonzalez

3.23

4.89

-1.66

Chen

3.54

3.89

-0.35

Norris

3.65

4.22

-0.57

Tillman

3.34

4.01

-0.67

Note: A negative difference indicates the pitcher "over-performed".

First, the fact that Jimenez can be considered to have under-performed his peripherals makes me question the entire idea of FIP, but I digress. Also, when you look at Gonzalez's difference - that was the biggest over-performance in the majors last year.

There was a good article written at the end of last season by Neil Weinberg at Fangraphs that examined the biggest ERA/FIP differences in 2014. There was only one problem - for some reason, Neil said it was the Nationals' Doug Fister with the biggest over-performance at -1.52. I'm not totally sure why Neil didn't include Miguel Gonzalez's numbers, but a lot of the analysis in that article could be indirectly applied here. And while Gonzalez did put up the "best" over-performance of last year, it's not that far off of Doug Fister. What about the rest of the league?

I used Baseball Reference to create a spreadsheet of all pitchers in the majors last year. I filtered out everyone under 75 IP to knock out small sample sizes. This left 190 cases total to examine (Note: This may include multiple partial seasons by a pitcher if he pitched for more than one team, but there isn't any crossover in the data). Of those, 109 (57%) over-performed their FIP.

The average difference (absolute value) between ERA and FIP for those 190 cases was 0.52 with a standard deviation of 0.43. And those numbers held up almost exactly when I looked at just those pitchers who did over-perform. So while Gonzalez's -1.66 is certainly an outlier, the other guys are about in line with the average difference last year. In other words, I wouldn't say there was anything about their performance to lead you to believe they're going to perform drastically different this year. It can get more complex if you look at each individual pitcher and his entire career, but a lot of these guys are young and without a ton of consistent pitching experience to begin with. And there's always that good defense there to back them up this year, just like 2012, 2013, and 2014.

As for Gonzalez, I did a search for similar numbers - an ERA between 3.2 and 3.3 and a FIP between 4.8 and 4.9 with at least 75 innings pitched. It's happened twice before in baseball history: Rich Rodriguez in 1991 and Al Benton in 1938. Both out-performed their FIP over the course of their careers, albeit with some varying results from year to year. Neither did so with the current magnitude of Gonzalez though, who posts a career difference of -1.14. I know I've spent this entire article questioning some of the validity of FIP in projection models, but that sure seems like a big number that's going to have to see some regression at some point.

The Orioles have made a habit of out-performing their projections, and while its impossible to tell after one game (albeit, a good one for the team) to know if they're going to do it again, if you want some kind of reason to think they may do it again in 2015 it could be that those projections may be under-estimating the Orioles' pitching and over-stating how lucky they've been in the past.