Bill James is not a fan of groundball pitchers. This is not new news; he’s written about them in the past on his site, Bill James Online. His most recent thoughts on the subject came last month in an essay entitled Two Bits, Four Bits. He addressed four separate topics:
1. The oddity of teams’ no. 1 starter being referred to as “not a true number one starter” when one never hears, say, a cleanup hitter being referred to as “not a true cleanup hitter”
2. The value of groundball pitchers vs. flyball pitchers
3. Whether facing a knuckleball pitcher screws up opposing hitters’ timing in the following game
4. How the ascendancy of Donald Trump indicates a challenge for the Republican Party
I’ll give you three guesses as to which topic drew the vast majority of the 64 comments that the post generated.
I’m going to focus on the second point. He ranked the 200 pitchers who faced the most batters since 1950 by groundball tendencies and concluded,
You don’t want ground ball pitchers. Ground ball pitchers suck. Ground ball pitchers are Jamey Wright and Mike Morgan. Fly Ball pitchers are Sandy Koufax and Dennis Eckersley…If you like Ground Ball pitchers, you’re welcome to them. I don’t want nothin’ to do with them.
Well, I thought, this goes against sabermetric orthodoxy, doesn’t it? I mean, in the BP Annual, groundball percentage is one of the 19 stats listed for every pitcher. Flyball percentage isn’t. And valuing groundball generation isn’t a new thing. So what are we all missing?
I looked at every pitcher from the 2015 season and ranked them by groundball tendencies, i.e., groundballs as a percentage of balls in play. I then divided them into deciles based on the number of batters they faced. Major-league pitchers faced 183,627 batters last year, so each decile consisted of 18,362.7 plate appearances. (There were nine pitchers who straddled two deciles; I divided their numbers proportionately.) Then I looked at some basic performance indicators for each group. Here are the results:
Decile |
BB% |
K% |
HR% |
DP% |
|||
1 |
60.5% |
3.45 |
3.59 |
7.9% |
19.7% |
1.8% |
15.4% |
2 |
52.7% |
3.84 |
3.81 |
8.1% |
19.9% |
2.2% |
13.7% |
3 |
49.8% |
4.01 |
3.98 |
7.7% |
18.8% |
2.4% |
12.6% |
4 |
47.7% |
3.76 |
3.70 |
7.1% |
20.6% |
2.4% |
11.7% |
5 |
46.0% |
3.65 |
3.83 |
7.6% |
20.7% |
2.5% |
12.4% |
6 |
44.2% |
4.15 |
3.98 |
8.4% |
20.8% |
2.6% |
10.5% |
7 |
42.3% |
3.84 |
3.88 |
6.7% |
20.9% |
2.8% |
10.3% |
8 |
40.1% |
4.44 |
4.26 |
7.9% |
20.1% |
3.1% |
10.8% |
9 |
37.6% |
4.34 |
4.23 |
7.5% |
21.7% |
3.4% |
9.6% |
10 |
31.8% |
4.18 |
4.38 |
7.7% |
20.6% |
3.5% |
8.4% |
Correlation |
(0.76) |
(0.88) |
0.23 |
(0.59) |
(0.97) |
0.97 |
Let’s go through those column-by-column. The last row lists the correlation coefficient between groundball percentage and the measure in each column.
In 2015, groundball pitchers in aggregate generated lower ERAs than flyball pitchers in aggregate. The correlation coefficient, -0.76, is pretty strongly negative. That’s even more true of FIP. Both measures of run prevention favor groundball pitchers.
Groundball pitchers allowed a few more walks than flyball pitchers, though the relationship wasn't particularly strong. They also got fewer strikeouts than flyball pitchers, and the relationship there is stronger.
Groundball pitchers are much, much better at preventing home runs. The -0.97 correlation coefficient represents an almost perfectly inverse relationship—the more flyballs allowed, the more home runs allowed. This, of course, is one of the most obvious advantages of groundball pitchers. It’s pretty hard hit a groundball out of the park.
None of this is particularly surprising. It confirms prior research, rather than blazes a new trail. So, I thought, in light of James’ comments, maybe I had something wrong?
One limitation is that James defined groundball pitchers by their ability to get double plays. I used groundball percentage. So maybe the definition that James used is at issue?
No, that’s not it. The last column of the table lists pitchers’ grounded into double plays as a percentage of opportunities (i.e., runner on first, fewer than two out). The correlation is almost perfect—groundball pitchers, as defined by groundball percentage, get a linear increase in GIDPs. So using GIDP percentage is a good proxy for inducing groundballs. It’s not a matter of James’ metric.
But in his article, James didn’t write about aggregate figures. He wrote about individual pitchers. The top six groundball pitchers, defined by GIDPs as a percentage of GIDP opportunities, were, alphabetically, Mike Caldwell, Scott Erickson, Whitey Ford, Mike Hampton, Tommy John, and Mel Stottlemyre. The bottom six—here he looked at GIDPs per batter faced rather than as a percentage of opportunities—were, alphabetically, Jim Bunning, David Cone, Eckersley, Catfish Hunter, Koufax, and Pedro Martinez. No question, the latter group’s better.
So let’s examine the top performers (measured by PWARP, BP’s DRA-based measure of pitcher wins above replacement) in each of the groundball deciles calculated above. In the top decile, the best pitchers in 2015 were Jake Arrieta, Dallas Keuchel, and Felix Hernandez. The second decile is headed by Carlos Carrasco, Sonny Gray, and Kyle Hendricks. Hmm. Of those six, one guy’s still great, two are hurt, and three are among the biggest surprises of the year so far, one good and two bad.
The third decile leaders are Clayton Kershaw, Jon Lester, and Hisashi Iwakuma. We’re still well into groundball territory here, and we’ve got the best pitcher in baseball and one of the best so far this year. The top 2015 performers in the fourth decile are Zack Greinke, Cole Hamels, and Masahiro Tanaka. In the fifth decile, it’s Chris Archer, Matt Harvey, and Noah Syndergaard. The sixth decile’s got Jacob deGrom, Danny Salazar, and Robbie Ray. Ray hasn’t been great but the others have been good, with some slow starters picking up the pace.
The seventh decile is getting into flyball territory, and it’s headed by Corey Kluber, Chris Sale, and Madison Bumgarner. No slouches there. The leading pitchers in the eighth decile are David Price, then a pretty big gap down to Wei-Yen Chen and Trevor Bauer. The ninth has Max Scherzer, Jake Odorizzi, and Mike Fiers. And the 10 percent of pitchers who yielded the fewest grounders are led by Colby Lewis, who’s been okay, Marco Estrada, who’s been good, and Dan Haren, who’s written some really funny tweets in retirement.
Looking at that list of top pitchers, you can sort of see James' point. Here are the 15 best pitchers from last year, by PWARP, with their groundball percentage decile in parenthesis. Remember, the lower the decile, the more groundballs:
1. David Price (8)
2. Clayton Kershaw (3)
3. Max Scherzer (9)
4. Corey Kluber (7)
5. Chris Sale (7)
6. Chris Archer (5)
7. Madison Bumgarner (7)
8. Zack Greinke (4)
9. Jake Arrieta (1)
10. Johnny Cueto (7)
11. Dallas Keuchel (1)
12. Carlos Carrasco (2)
13. Colby Lewis (10)
14. Jacob deGrom (6)
15. Jake Odorizzi (9)
The mean decile ranking is 5.7. That means that of the top 15 pitchers, there was a slight bias in favor of flyball pitchers compared to groundball pitchers. If you weight the decile ranking by PWARP ranking, the skew toward flyball pitchers increases.
But this appears to be a case of losing sight of the forest for the trees, doesn’t it? There are some really great flyball pitchers. But as the table above shows, groundball pitchers, in aggregate, outperform flyball pitchers, in aggregate.
That isn’t to say that an individual groundball pitcher is better than an individual flyball pitcher. Among ERA qualifiers, the most extreme groundball pitcher last year was Brett Anderson, who had a 3.69 ERA, 3.97 FIP, 6.71 (!) DRA, and 101 ERA+. His 66.3 percent groundball rate generated results that were, at best, league average. By contrast, among pitchers in the tenth decile, producing the fewest groundballs, were the three starters listed above as well as relief studs Kenley Jansen (2.41 ERA, 2.17 FIP, 2.18 DRA), Darren O’Day (1.52/2.46/2.40), Roberto Osuna (2.58/2.99/2.60), and Cody Allen (2.99/1.79/2.23). So while flyball pitchers, in aggregate, underperform groundball pitchers, that doesn’t say that individual flyball pitchers aren’t elite. So the answer appears to be, as in so much analysis, it depends.
But as far as wanting nothin’ to do with groundball pitchers, well, thus far in 2016, among ERA qualifiers, Jake Arrieta is fourth in groundball percentage, Steven Matz is fifth, Aaron Nola is 16th, Noah Syndergaard is 17th, Clayton Kershaw is 21st and Danny Salazar is 23rd. I want anythin’ to do with them, as I do with Justin Verlander (fifth lowest groundball percentage), and Max Scherzer (seventh lowest), Jose Quintana (12th) and Chris Sale (20th).
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.
Subscribe now
If you were starting a playoff series which of those pairs would you take? I'd take the latter pair right now.
(looks over shoulder, nervously searches for "Kershaw has a 7.23 ERA in the NLCS" comment)
"Groundball pitchers got fewer strikeouts than flyball pitchers, though the relationship wasn’t particularly strong. The tradeoff is that groundball pitchers were better at preventing walks, and the relationship there is stronger."
Walks have only a 0.23 correlation, while Ks have a -0.59 correlation.
I think the real takeaway is that GB/FB may be a way to explain pitchers' results, but trying to use it as a filter to find the pitchers you want is a bad idea. Commit to only flyballers, and the back-end of the rotation is a disaster. Commit to only groundballers, and you miss out on the greats.
Sounds more like he's arguing about uniquely effective pitchers as a class. Not groundballers in particular. Not to mention that the more guys you strike out the better ;-)
As a result, many of the very best pitchers in baseball are flyball pitchers. But, on average, you're betting off keeping the ball on the ground. The variations in strikeouts and walks are mostly window-dressing around that core dynamic.
FIP contribution by decile, HR/BB+HBP/K (positive numbers bad, negative numbers good):
1: 1.00/1.11/-1.65
2: 1.22/1.14/-1.68
3: 1.34/1.11/-1.60
4: 1.30/1.01/-1.74
5: 1.38/1.05/-1.73
6: 1.44/1.19/-1.78
7: 1.53/0.97/-1.76
8: 1.72/1.13/-1.72
9: 1.87/1.06/-1.85
10: 1.92/1.08/-1.75
Total variance in FIP contribution: HR 0.92, BB+HBP 0.22, K 0.25. Throw out the high/low outliers and it's HR 0.67, BB+HBP 0.18, 0.10. So your intuition is correct: The variance in HRs is the primary driver. So if you get an FB pitcher who limits the longball, that's going to be an extremely effective pitcher. But that's also not the norm.