June 22, 2016

Prospectus Feature

Groundballs Through the Decades

by Rob Mains

I am a Bill James fan. You might not be. That’s okay. For my money, if there’s no Bill James, I’m probably not writing this.

So it caused some dissonance on my part when I wrote that I thought James’ antipathy for groundball pitchers is misplaced. One of my general rules in life is that if you come to a conclusion that’s seriously at odds with conventional thinking, you could well be right, but to be safe, be sure you can answer in the negative the question Am I doing something stupid?

I really don’t think I am. I found that groundball pitchers may—may—allow a few more walks, and they get a handful fewer strikeouts, but by allowing a lot fewer homers, they allow fewer runs. That’s true in terms of both outcomes (ERA) and process (FIP). Here’s the relationship:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	61.8%	3.47	3.62	7.9%	19.4%	1.9%
2	54.2%	3.95	3.81	8.0%	19.6%	2.2%
3	51.6%	3.92	3.94	7.7%	19.2%	2.4%
4	49.4%	3.73	3.72	7.5%	20.9%	2.3%
5	47.7%	3.67	3.83	7.2%	20.5%	2.6%
6	46.1%	4.16	4.01	8.4%	20.5%	2.6%
7	44.1%	3.96	3.97	7.2%	20.6%	2.8%
8	41.9%	4.25	4.12	7.3%	21.0%	3.1%
9	39.5%	4.30	4.18	7.6%	21.5%	3.2%
10	33.5%	4.22	4.43	7.7%	20.6%	3.6%
Correlation		(0.80)	(0.92)	0.27	(0.75)	(0.98)

I ranked every 2015 pitcher by groundball rate, and divided them into 10 equal-sized deciles based on plate appearances. As you can see, the pitchers who yielded the most grounders on batted balls averaged 61.8 percent grounders. They gave up homers in 1.9 percent of plate appearances, compiling an aggregate ERA of 3.47 and FIP of 3.62. By contrast, the pitchers who gave up the fewest grounders, averaging 33.5 percent, had a 3.6 percent home run rate, a 4.22 ERA, and a 4.43 FIP. That’s pretty clear, right? And those aren’t cherry-picked numbers. ERA, FIP, and home run rate are all negatively correlated to groundball rate, as the last line shows. The more grounders, the fewer homers and the lower ERA and FIP. The regression relationship is this: For every 10 percentage point increase in groundball rate, pitchers reduce their ERA by 0.28, FIP by 0.27, and home run percentage by 0.006 percent. So James is wrong when he says, “If you like Ground Ball pitchers, you’re welcome to them.” Isn’t he?

Except. That’s not exactly what James was saying. I mean, he wrote it, but he wasn’t referring just to the 2015 season. He was looking at the 200 pitchers who faced the most batters from 1950 to the present. That’s not precisely the same thing. Maybe there’s something about groundball pitchers’ career value that’s different from the snapshot that 2015 provides.

I retrieved data for every pitcher who’s logged at least 1,000 innings since 1950. That gave me 708 pitchers. Granted, that’s a much larger sample than James’ 200, but if the results hold for 200 pitchers, they should for 708. And James used groundball double play percentage (GIDPs divided by double play opportunities, i.e. a runner on first and fewer than two out) as his proxy for groundball percentage. So I did the same. My data ranged from Chris Young (5.7 percent GIDP rate in 1,179 innings) to Zane Smith (18.3 percent GIDP rate in 1,919 innings).

So are the results similar to the table above? Here’s what I found:

Decile	GIDP%	ERA	FIP	BB%	K%	HR%
1	14.9%	3.83	3.87	7.4%	12.2%	2.0%
2	13.1%	3.84	3.78	7.3%	14.0%	2.1%
3	12.3%	3.81	3.85	7.6%	13.5%	2.1%
4	11.6%	3.86	3.87	8.0%	14.4%	2.2%
5	11.0%	3.70	3.80	7.7%	14.5%	2.2%
6	10.6%	3.85	3.81	7.8%	16.0%	2.4%
7	10.1%	3.81	3.81	8.0%	16.1%	2.4%
8	9.6%	3.79	3.77	8.1%	16.5%	2.4%
9	8.9%	3.74	3.78	8.4%	16.9%	2.4%
10	7.7%	3.69	3.72	8.9%	18.6%	2.4%
Correlation		0.63	0.75	(0.91)	(0.97)	(0.91)

Yikes. It’s broken down. For pitchers with at least 1,000 innings, getting a higher percentage of GIDPs is associated with fewer walks, a lot fewer strikeouts, and fewer home runs…but a slightly higher ERA and FIP. How’s that?

Maybe it was James’ use of GIDP percentage as his groundball metric. The major online databases have varying availability of groundball data. Here at BP, our groundball data extends all the way back to 1950 (though with less reliability), so I was able to retrieve groundballs as a percentage of batted balls for each of the 708 pitchers with 1,000 innings pitched since 1950. They range from Young again (28.7 percent groundball rate) to the pitcher James calls Surgery (68.6 percent). Here are the data:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	61.5%	3.56	3.65	7.5%	12.9%	1.7%
2	56.1%	3.72	3.76	7.9%	13.7%	2.0%
3	52.8%	3.75	3.78	7.9%	14.2%	2.1%
4	50.7%	3.72	3.81	7.8%	14.1%	2.1%
5	49.3%	3.68	3.72	7.9%	15.7%	2.2%
6	48.0%	3.93	3.85	8.2%	15.8%	2.3%
7	46.5%	3.89	3.92	8.3%	15.6%	2.4%
8	45.0%	3.82	3.76	7.7%	16.5%	2.5%
9	43.2%	3.99	3.87	8.0%	16.8%	2.6%
10	39.2%	3.88	3.91	7.9%	17.3%	2.8%
Correlation		(0.82)	(0.79)	(0.50)	(0.96)	(0.99)

There, that’s more like it. More groundballs means fewer homers, fewer strikeouts, fewer walks (maybe—that’s not much of a correlation), and lower ERA and FIP.

And yet…In 2015, the spread between the highest and lowest deciles for ERA was 0.99 runs. The spread for FIP was 0.79 runs. For pitchers with 1,000 innings since 1950, the spreads are much tighter, 0.43 and 0.27, respectively. The correlation’s negative—more grounders mean fewer runs allowed—but the effect is much more muted than in 2015. In 2015, gaining 10 percentage points of groundball percentage was related to a 0.28 reduction in ERA and a 0.27 reduction in FIP. For the 1950-present cohort, the reduction’s much lower, 0.17 for ERA and 0.10 for FIP. So I’m not giving up on Bill James yet! That 10th decile above includes Hall of Famers Sandy Koufax (38 percent groundball rate), Dennis Eckersley (39 percent), Early Wynn (40 percent), Catfish Hunter (40 percent), and Robin Roberts (41 percent). The only Hall of Famers in the top decile are Bruce Sutter (62 percent) and Greg Maddux (59 percent).

Note, though, that those 10th-decile Hall of Famers all ended their careers before the turn of the millennium. Specifically, Eck threw his last pitch in 1998, Cat in 1979, and the other three in the 1960s. Maybe this is something that’s changed over the years—perhaps groundball pitchers have become more effective in recent years. The 2015 data clearly show an advantage for groundballers. Is it a constant?

I checked six other years, going back a decade at a time. Here’s 2005:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	60.0%	3.83	3.97	7.7%	15.2%	2.1%
2	52.7%	4.17	4.18	8.6%	16.1%	2.3%
3	50.2%	4.09	4.18	7.7%	15.3%	2.5%
4	47.6%	4.02	4.02	7.6%	17.2%	2.5%
5	46.1%	4.31	4.24	8.0%	16.0%	2.6%
6	44.4%	4.38	4.35	8.1%	16.7%	2.8%
7	42.4%	4.48	4.43	8.4%	16.9%	2.9%
8	40.6%	4.30	4.32	8.2%	16.5%	2.8%
9	38.2%	4.57	4.60	8.6%	17.2%	3.2%
10	33.4%	4.78	4.63	8.7%	17.3%	3.3%
Correlation		(0.92)	(0.90)	(0.62)	(0.80)	(0.97)

That looks a lot like 2015. Each 10 percentage point increase in groundball percentage correlates to a 0.34 reduction in ERA and 0.26 reduction in FIP. The correlations are strong. Groundballers rule.

Here’s 1995:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	60.7%	4.03	4.05	8.6%	15.8%	1.9%
2	54.8%	4.24	4.30	9.0%	15.0%	2.1%
3	52.4%	4.44	4.42	8.7%	14.6%	2.4%
4	50.3%	4.43	4.51	9.2%	14.8%	2.5%
5	48.6%	4.50	4.30	8.6%	16.2%	2.5%
6	46.9%	4.17	4.11	8.2%	17.9%	2.5%
7	44.6%	4.68	4.58	9.3%	16.6%	2.8%
8	41.9%	4.37	4.50	9.3%	17.3%	2.8%
9	39.4%	4.54	4.75	9.6%	16.0%	3.0%
10	34.1%	5.13	5.02	10.4%	18.2%	3.6%
Correlation		(0.80)	(0.84)	(0.73)	(0.67)	(0.98)

Two things about 1995: First, it was a strike-shortened year. There were only 144 games per team. Second, there were 4.85 runs per game in 1995, the seventh most since World War II. But since that total was lower than in the surrounding years of 1994 (4.92 in another strike-shortened year) and 1996 (5.04), I figured it was okay to go with 1995, short season or no. And, as you can see, the results are consistent with the correlations and spreads in 2005 and 2015. A 10 percentage point increase in groundball percentage lowers ERA and FIP by 0.31.

Now 1985, the year that still resonates for replay review advocates:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	68.9%	3.82	3.78	9.4%	12.4%	1.6%
2	61.2%	3.88	3.76	8.2%	13.3%	2.0%
3	57.3%	3.97	3.86	8.8%	13.1%	2.0%
4	54.3%	3.90	3.84	7.5%	12.6%	2.2%
5	52.4%	3.75	3.90	9.0%	14.2%	2.2%
6	50.3%	3.97	3.86	8.4%	14.5%	2.3%
7	48.6%	3.83	3.77	8.7%	14.9%	2.1%
8	45.9%	3.68	3.85	8.5%	14.6%	2.3%
9	43.4%	4.18	4.19	8.6%	14.5%	2.9%
10	36.7%	3.95	4.12	9.2%	16.0%	2.9%
Correlation		(0.25)	(0.71)	0.00	(0.89)	(0.93)

Things look a little different here. Getting a lot of groundballs in 1985 was less correlated to run prevention than it was in 1995, 2005, or 2015; a 10 percentage point increase in groundball percentage yielded only a 0.04 reduction in ERA and a 0.11 reduction in FIP. It was worth over a quarter of a run in the prior years.

1975, the first year of the Big Red Machine’s back-to-back Series wins:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	65.2%	3.68	3.67	8.5%	11.4%	1.6%
2	59.1%	3.67	3.72	8.9%	11.0%	1.5%
3	56.5%	3.85	3.64	9.0%	11.8%	1.5%
4	54.3%	3.89	3.80	9.3%	13.0%	1.9%
5	52.6%	3.77	3.71	9.2%	12.9%	1.7%
6	50.8%	3.58	3.78	9.3%	12.6%	1.8%
7	48.6%	3.61	3.58	8.9%	15.1%	2.0%
8	46.8%	3.68	3.63	8.7%	14.2%	2.0%
9	44.9%	3.76	3.70	8.8%	14.3%	2.1%
10	39.6%	3.62	3.88	9.5%	13.4%	2.1%
Correlation.		0.26	(0.31)	(0.47)	(0.77)	(0.87)

Have you ever run a birthday party for little kids, like 6-year-olds? For a while things go pretty smoothly. You have activities planned, and everything goes well. The kids do as they’re told and everybody’s happy. This groundball/flyball comparison here is the point in the party where a couple kids stop coloring and instead are running around, and one’s crying for no apparent reason, and one—well, you don’t know where one kid is. Extreme flyball pitchers, the tenth decile, now have the lowest ERA; they were among the highest in the other four seasons. For the first time so far, they don’t have the highest strikeout rate. And there is no discernible relationship between groundball rate and FIP. A 10 percent increase in groundball rate now yields a 0.04 increase in ERA, a 0.04 decrease in FIP, and a tiny 0.02 percent reduction in home run rate. Groundball superiority is breaking down.

Here’s 1965, the year the Yankee dynasty died:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	67.1%	3.20	3.36	8.5%	14.3%	1.7%
2	60.6%	3.67	3.47	8.5%	15.0%	1.9%
3	58.5%	3.29	3.44	7.3%	14.4%	2.1%
4	55.7%	3.41	3.22	8.2%	16.1%	1.8%
5	52.8%	3.40	3.50	7.7%	15.9%	2.3%
6	50.5%	3.54	3.66	8.3%	14.6%	2.3%
7	48.7%	3.82	3.51	8.3%	16.2%	2.2%
8	46.8%	3.35	3.53	8.4%	17.4%	2.4%
9	44.2%	3.63	3.55	8.4%	17.5%	2.5%
10	40.2%	3.70	3.75	8.7%	16.9%	2.8%
Correlation		(0.58)	(0.70)	(0.24)	(0.80)	(0.92)

You’ve just brought out the birthday cake. Every kid is sitting at the table, eating cake and drinking juice. You’ve re-established order. Extreme groundball pitchers have the lowest ERA, second-lowest FIP, lowest strikeout rate, lowest home run rate. Extreme flyball pitchers have the second-highest ERA, highest FIP, third-highest strikeout rate, highest home run rate. The correlations are back where they were in other years. Getting more grounders has a smaller impact on run production than in the recent past—a 10 percentage point increase in groundball percentage results in a 0.14 reduction in ERA and a 0.13 reduction in FIP—but at least the sign is correct.

And 1955, the year Brooklyn finally won a World Series:

Decile	GB%	ERA	FIP	BB%	K%	HR%
1	61.3%	3.97	3.86	9.4%	10.7%	2.3%
2	56.7%	4.19	4.04	10.1%	11.2%	2.5%
3	53.9%	3.97	3.90	9.7%	11.1%	2.3%
4	51.5%	4.33	4.08	10.8%	11.5%	2.3%
5	48.1%	3.99	4.19	9.2%	11.1%	2.9%
6	46.1%	3.94	4.02	10.9%	12.8%	2.5%
7	44.6%	4.22	3.96	11.1%	12.4%	2.3%
8	42.5%	3.73	3.86	10.5%	13.6%	2.3%
9	39.5%	3.86	3.97	9.6%	13.6%	2.7%
10	34.5%	3.79	4.12	10.1%	13.4%	2.9%
Correlation		0.50	(0.29)	(0.25)	(0.90)	(0.54)

All that sugar’s gotten the kids totally revved up, everybody’s running around, one’s trying to ride the dog like a horse, one’s managed to lock herself in the bathroom, and where the hell are all the parents to pick up these brats? Getting more groundballs is, as in 1975, correlated with a higher ERA, but it’s worse: 0.12 per 10 percentage points of groundball percentage. The relationship between groundballs and FIP looks pretty random. Even the correlation between groundballs and home runs, the most intuitive one on these tables, is breaking down.

Assuming that groundball data from the far-back past are reliable enough for us to trust these tables, there are two conclusions, I think. First, I’m not backing away from the view that in contemporary baseball, groundball pitchers, in aggregate, are more valuable the flyball pitchers, in aggregate. As pointed out above, groundball pitchers’ relative paucity of strikeouts is more than outweighed by their stinginess with the longball, making them better at preventing runs.

But those relationships aren’t static. The clear advantage of groundball pitchers hasn’t been constant. So when James talks about pitchers over a 65-year stretch, sure, there are a lot of really successful flyball pitchers. That’s skewed toward earlier years, when groundball pitchers didn’t enjoy the same advantage that they do today, but it doesn’t invalidate his view.

The second conclusion is that we as analysts need to keep in mind how much the game’s changed over the years. To be honest, I struggle to explain how extreme flyball pitchers, who are, in aggregate, less effective than extreme groundball pitchers in today’s game, were able to completely reverse that deficit 40 or 60 years ago. Joe Sheehan often points out that baseball 50 years ago was very different than it is today. When I hear that, I think, “Well, sure, more strikeouts and more relief pitchers today,” but it goes beyond that. Basic tenets shift over time. Baseball changes more than we think.

Rob Mains is an author of Baseball Prospectus. Follow @Cran_Boy
Click here to see Rob's other articles. You can contact Rob by clicking here