The Workloads of the Top College Pitchers in the 2018 Draft

(via Michelle Jay)

A couple of weeks before tonight’s Major League Baseball draft, University of Kentucky starter Sean Hjelle was pushed hard. On May 17, the 6-foot-11 right hander threw 123 pitches in an important game against rival Vanderbilt. He dealt with jams all through his six innings—averaging two baserunners per frame—and allowed four runs en route to an 8-1 loss. The Wildcats’ conference record dropped to 13-16, raising the stakes for the following week’s SEC tournament. To earn a coveted spot in Regionals, Kentucky needed to advance in this single-elimination conference competition.

In their May 22 SEC opener against Auburn, Kentucky fell into a 3-1 hole in the third inning, and the Tigers continued to threaten. Needing to stop Auburn from scoring further, Wildcats head coach Nick Mingione went back to his ace. Hjelle entered the game in relief and threw 7.1 scoreless innings, allowing four hits, striking out nine and walking none. Just five days after his stressful start, Hjelle stayed in this game for an eye-opening 118 pitches.

Rarely does a modern big league manager ask a pitcher to work as hard as Hjelle did. In the past five seasons, just four MLB pitchers have had consecutive games of at least 118 pitches on four days of rest or less. Not only were those MLB hurlers more developed physically, but they all were guaranteed multimillion-dollar salaries. Hjelle, meanwhile, was going unpaid and running a risk. An all-out pursuit of these collegiate victories could come at the expense of ever getting a payout for his pitching ability—whether the overuse resulted in a deflated signing bonus or a higher probability of injury to dampen future earnings potential.

Of course, two games alone don’t provide a clear picture of whether overuse has been an ongoing problem for Hjelle; we need to weigh the full slate of games over his Kentucky career. And just like Hjelle is an important name in this year’s draft—FanGraphs recently projected that he’ll be selected early in the second round—the same can be said for the skyscraping righty’s peers: the top draft-eligible collegiate pitching talents of 2018. Many of these pitchers will be selected tonight and ascend organizational prospect rankings the moment they sign with their drafting clubs. These soon-to-be minor leaguers will foster hope, and it’s important that we assess whether they’ve been used heavily in college—as is the case for so many pitchers across NCAA baseball.

Set-up

We’ll focus on the best ten college pitchers, as ranked by FanGraphs’ Eric Longenhagen and Kiley McDaniel.

To gather pitch counts for this group of ten, I pulled the full 2016–2018 game logs (pre-NCAA tournament) from their college statistics pages. When pitch counts were missing, I consulted box scores from team schedules, which usually contained the omitted tallies. After that process was complete, just 16 of the group’s 449 total appearances were missing pitch counts. Even though that’s a slim rate of null values (just 3.6 percent of games), these need to be filled for the analysis to be completed in full.

In the past, I’ve estimated pitch counts by multiplying the total batters faced in a game times a pitcher’s average number of pitches per batter faced. This year, I added rigor to the interpolation effort. I trained a gaussian GBM model based on daily box score stats: innings pitched, hits, runs, earned runs, walks, strikeouts, and batters faced. Additionally, controls for the pitcher identity and year were included.

In testing, the mean absolute error of the model was +/– seven pitches. While that error rate could push borderline pitch counts over certain usage thresholds, it certainly won’t radically alter our impression for any of the 16 interpolated appearances. With those solid estimates and all the actual pitch counts already under our belts, we can assess how heavily these pitchers have been used.

How Their Workloads Stack Up

First, let’s get an idea of the group’s game-by-game pitch counts. Shown below are ridgeline–style density curves depicting the pitch counts amassed over the 2016–2018 period.

Starting from the left, we see most of these pitchers—particularly Bubic, Gilbert, Hjelle, Rolison, and Singer—saw their fair share of short outings. Early in their college careers, they spent significant time in the bullpen. Only Lynch, McClanahan, and Pilkington were essentially able to slot right into their teams’ rotations as freshmen.

But that doesn’t mean the young pitchers were treated with kid gloves all through college. If we look rightward, we see every distribution peaks in starter pitch-count territory, albeit in different spots. The most frequent pitch counts posted by Hjelle, Kowar, McClanahan, Mize, and Rolison were under 100 pitches, which is a positive for MLB teams concerned about heavy usage. Bubic’s peak was right around the century mark, while the maxima from Lynch, Pilkington, and Singer bled over the 100-pitch threshold.

The high-usage standout, though, is Gilbert, as his distribution peaks at 107 pitches. That farther apex suggests the Stetson righty was worked harder than the others in this talented group. Gilbert’s distribution is also notable for seeming to have the most area in high-usage territory, at pitch counts between 120 and 140.

Let’s pin down precisely the frequency of high-pitch outings for Gilbert and the rest. The heat map below shows the portion of appearances in which each pitcher threw at least 115 pitches. For reference with these numbers, it’s worthwhile to know that my 2016 investigation of college pitcher usage revealed regular-season starts spanned 115+ pitches for freshmen 3.6 percent of the time, for sophomores 7.2 percent of the time, and for juniors 9.9 percent of the time. In tournament play, these rates roughly double.

Half of the seasons shown don’t have any starts crossing the 115-pitch threshold. But those that do stick out. As is often the case in the NCAA, the workloads for Gilbert and Lynch grew progressively heavier each year. As juniors, they topped 115 pitches in an excessive ~30 percent of their games. Gilbert’s workload is particularly glaring upon even closer review. This year, he has had three starts of at least 122 pitches, reaching a high of 128 in an eight-inning effort against Lipscomb University in April. And last season, Gilbert had a 135-pitch start, the lengthiest outing from any of the top ten pitchers.

Not to be ignored are a few other high-intensity seasons revealed in the heat map. Hjelle eclipsed the NCAA norms in both his sophomore and junior seasons, which isn’t a surprise given the workload he shouldered in May. And one-fifth of Bubic’s sophomore starts surpassed the 115-pitch mark, which is more than twice the average 2012–2015 rate for second-year college pitchers.

These pitch-count-centric looks are useful for benchmarking. We would also do well to factor in rest, the other vital element of usage. Let’s incorporate MLB’s Pitch Smart guidelines, which are a reasonably objective standard for evaluating whether pitchers are getting enough recovery time after their individual outings. Assembled by a host of doctors and medical experts, the Pitch Smart recommendations for college-aged pitchers are as follows:

While the guidelines are purposely general—acknowledging that a truly safe innings limit varies from arm to arm—they’re useful because they allow us to flag outing-rest pairings that are excessive for the typical pitcher. The next heat map illustrates the rate of appearances that register as “violations” of these guidelines.

Here, over half of the group’s seasons don’t pick up any violations, which is a positive development for the players and the drafting clubs. Of particular note in that crowd of zeroes is Shane McClanahan, who had Tommy John Surgery in 2016 but was not worked to excess upon his return to South Florida’s rotation.

We do see boxes notable for high violation rates. Hjelle, Pilkington, and Singer each had a season that didn’t follow Pitch Smart in at least 10 percent of appearances. And importantly, Gilbert’s results signal overuse in this area as well; the Pitch Smart violation rate for the Hatters ace ballooned to 21 percent in 2018.

Concluding Remarks

If Gilbert is selected in the first round tonight, there’s no doubt his new club would be getting a high performer. Before the NCAA tournament began, he had posted a sterling 2.52 ERA and 37.3 percent strikeout rate, outperforming Casey Mize (the likely No. 1 selection) in the process. Gilbert’s control was strong, too, as his 7.15 K/BB was one of college baseball’s top ratios. Plus, he brings solid pitcher characteristics to the table: He shows promising command and athleticism, and his frame—at 6-foot-6, 225 pounds—seems built to handle a starter’s workload.

It’s possible those attributes give Gilbert a higher workload tolerance than other pitchers. But we can’t say for certain, particularly since overuse may have already taken a toll in 2018, when he brought diminished velocity to the mound for much of the season. Last year, Gilbert showcased mid-90s heat in the spring and summer with above-average off-speed stuff. With the heavy use in 2018, his fastball was in the 89–93 mph range with plainer secondaries. Even as his stuff has picked back up in his last few outings, there are some reasons for concern, and interested teams will need to decide whether a fresher arm might be a safer selection.

References and Resources

• The NCAA’s main game logs, and a host of individual college program schedules and boxscores
• MLB’s Pitch Smart guidelines
• THE BOARD
• Gerald Schifman, The Hardball Times, “How Hard are College Pitchers Worked?”
• Gerald Schifman, FanGraphs, “The Workloads of 2017’s Top Draft Prospects”