Estimating Fastball Groundball Rate and Swinging Strike Rate With Small Samples

The Royals recently promoted 2014 draftee Brandon Finnegan to the majors to be a lefty out of the bullpen. So far the 21-year-old has thrown just a couple of relief innings. What I wanted to do is push the limits to discover what type of pitcher Mr. Finnegan may be in the future given this small sample of information. Today, I am concentrating just on his fastball.

Eno has been doing quite a bit of pitch benchmarks and I helped him out with some values. After using these values for a few months, I noticed heavy flyball pitchers like Chris Young and Danny Duffy had vertical movements, as defined be Pitchf/x, near or above 10. On the other end of the spectrum, ground ball pitchers, like Justin Masterson, had a vertical component near or below zero. By knowing the amount of downward break on fastball, I hope to get a range on their ground ball rate (GB%).

Additionally, a fastball is thrown fast with the hope to induce some swing-and-miss. So, additionally I looked to find the average swinging strike rate (SwStr%) just by knowing a pitcher’s velocity. Since velocity stabilizes quickly, just a few pitches will give us an idea of the pitcher speed and hopefully the average amount of swing-and-miss from their fastball.

Goundball Rate

I looked at all pitcher seasons from 2010 to 2014 where the pitcher threw a 2-seam, 4-seam, sinker or cutter at least 100 times. Then, I used the horizontal break (found here on a player’s FG page, z-Mov) and linear regression to help find the groundball rate. I did the same for fastball velocity and swinging strike rate. We know that vertical break and velocity will not be the only factors determining the pitches outcome, but they are significant factors. Here is what I found.

The best results were with vertical break and GB%. The exact final equation I found for all four pitches is:

GB% = .598 – .026 * (Vert Break)

Additionally, one standard deviation from the mean is +/- 5%.

So for a general quick rule, start with 60% GB% and then subtract 2.5% for the amount of vertical movement over zero or add 2.5% for the amount under zero.

So going back to Finnegan, he had a 7.6 vertical movement, so 60% – 2.5%* 7.6 = 41% (+/- 5%). Looking back at the original benchmarks for a four-seam fastball, a 37% GB% is league average, so there is decent change Finnegan’s will also be league average.

Velocity

First, for cutters, velocity as NO relationship on SwStr%. None. Assume a SwStr% of 8.8%

For the other three pitches, here are the equations:

SI: .0015 * (Velo) – .0802
FF: .0037 * (Velo) – .2781
FT: .0032 * (Velo) – .2378

For quick reference, here is a table with all the values:

MPH SI FF FT
84 4.6% 3.3% 3.1%
85 4.7% 3.6% 3.4%
86 4.9% 4.0% 3.7%
87 5.0% 4.4% 4.1%
88 5.2% 4.8% 4.4%
89 5.3% 5.1% 4.7%
90 5.5% 5.5% 5.0%
91 5.6% 5.9% 5.3%
92 5.8% 6.2% 5.7%
93 5.9% 6.6% 6.0%
94 6.1% 7.0% 6.3%
95 6.2% 7.3% 6.6%
96 6.4% 7.7% 6.9%
97 6.5% 8.1% 7.3%
98 6.7% 8.5% 7.6%
99 6.8% 8.8% 7.9%
100 7.0% 9.2% 8.2%

For all useful purposes, the standard deviation for all three pitches was +/- 1.5%. The deviation is quite large, but again I am just looking for a general idea of what to expect.

A rule of thumb is a little harder to come up with, but this is the best I could do. Start with 5.5% SwStr% for 4-seamers and 5% for a 90 mph pitch. Then increase the SwStr% 0.33% for each 1 mph increase in velocity.

For Finnegan’s 4-seamer, it averaged 92.6 mph, so 5.5% + .33 * 2.6 = 6.4% SwStr% +/- 1.5%. For a 4-seam fastball, the projected value is just at league average, but the error range could push it higher or lower.

Additionally, here are the estimated values for a couple of other recent callups, Daniel Norris and Matt Barnes.

Daniel Norris
4-seamer estimates (10.4 vert move, 91.6 mph)
GB%: 34% +/- 5% (Average GB%)
SwStr: 4.9% +/- 1.5% (Below average SwStr%)

Matt Barnes
4-seamer estimates (11.5 vert move, 93.6 mph)
GB%: 31% +/- 5% (Below average GB%)
SwStr: 6.7% +/- 1.5% (Above average SwStr%)

While a quite a few factors go into how a pitch will eventually play out over time, we know that more downward break leads to more groundballs and higher velocity leads to more swings and misses. By looking at these two values for fastballs, an estimate for groundball rate and swinging strike rate can be determined. These values can then be used as brenchmarks going forward.





Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.

5 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Jim K
9 years ago

I’ve done some similar work to the problem you’re working on, using multi-variate regression to tease out what makes secondary pitches effective. My goal was to test some of the assumptions I had been making in trying to interpret the pitchf/x game charts on the player pages. Once the season is over, I’ll find it, update and put on Community Research. The model is surprisingly predictive.

Also, check out the game charts for Daniel Norris. Then pull up Clayton Kershaw’s. Eerily similar, especially the curveball. Helps to explain his crazy high K%’s in the minors and bodes well for major league success.

Jim Kelleymember
9 years ago
Reply to  Jeff Zimmerman

My previous tries have used SwStrk% on all pitches, rather than by pitch type. Those are shown on player pages here at FanGraphs, but not on the custom leaderboards (at least as far as I can tell). Where do you and Eno get your raw data from? It would be much easier & more precise to have all the plate discipline and batted ball data available for each pitch type, rather than relying on overall numbers.