Sabermetrics: The Astros Ground Game

Crash Davis: Relax, all right? Don't try to strike everybody out. Strikeouts are boring! Besides that, they're fascist. Throw some ground balls - it's more democratic.

That has to be one of the most famous quotes from "Bull Durham," right? The Astros' off-season pitching acquisitions seem to be following that advice. Free agent relievers Chad Qualls and Matt Albers are coming off of top 10 groundball rate seasons for the last two years. The Astros also signed groundball starting pitchers Scott Feldman and Jerome Williams to join rotation candidates Jerrod Cosart, Dallas Keuchel, and Lucas Harrell, who posted groundball rates above 50 percent.

Worm Burner specialists may be economical, as well as democratic. Arguably, groundball pitchers can provide a cheaper option in the free agent market for achieving run prevention. Flamethrowers who can strike out the side in the 9th inning will always be in high demand---and thus expensive. The groundball relievers are not likely to get nicknames like "Lights Out," but they can be effective at getting the inning ending double play.

Beyond the Boxscore's Ryan Morrison examined whether elite groundball rate relief pitchers are undervalued. Qualls and Albers were in the group of elite groundball rate pitchers he analyzed. Among his findings: (1) ERA and FIP may understate the value of these relievers; (2) These relievers tend to post RE24 performance which outpaces their basic peripherals; and (3) Standard pitching metrics like FIP and x-FIP do not reflect the higher ground into double play (GDP) rates achieved by the extreme groundball relief pitchers. His article discusses possible platoon strategies which can be used to target elite groundball relievers for high leverage situations. I won't repeat the discussion, but I encourage you to read his article.

Since his article addresses relief pitchers, I instead will turn to groundball starting pitchers. First, I will analyze the pitching characteristics of groundball starting pitchers. Second, I will examine the GDP performance of Cosart and Feldman, in particular. An example of a Jarred Cosart induced GDP is shown below.

Starting Pitchers By Groundball Rate, 2011 - 2013

For starting pitchers' performance over the 2011 - 2013 period, I grouped the pitchers into three categories: (1) High Groundball Rate (49% and over GB%); Moderate Groundball Rate (43% - 49% GB%); and Low Ground Rate (Below 43% GB%). The number of pitchers in each group are shown below:

	# Pitchers
HIGH	27
MODERATE	60
LOW	38

The high group can be viewed as an elite GB rate, while the moderate group consists of average to somewhat above average groundball rates. The low group is clearly below average in groundball rate.

		Starting Pitchers' Metrics 2011 - 2013
GB% Rate
	K/9	BB/9	HR/9	BABIP	LOB%	Avg. GB%	HR/Fly	ERA	FIP	x-FIP
HIGH	6.508	2.897	0.842	0.298	71.5%	53.4%	11.3%	3.99	3.92	3.82
MODERATE	7.505	2.672	0.947	0.295	72.9%	45.9%	10.6%	3.84	3.75	3.73
LOW	7.226	2.806	1.138	0.287	73.0%	38.3%	10.3%	4.05	4.12	4.14

We will attempt to separate the wheat from the chaff in this table. Some observations:

Strike outs: Not surprisingly, the high GB% pitchers have the lowest strike out rate. Sinkers and two seam fastballs are intended to induce contact. The moderate GB% group includes some of the elite pitchers in the game, such as Clayton Kershaw, Max Scherzer, and Justin Verlander, which likely accounts for the group's superior K rate over the low GB% pitchers.

HR/9: The most significant attribute of groundball pitchers is the ability to suppress home runs. The higher the GB% tier, the lower the HRs per nine innings.

BABIP: The average BABIP for the period is .294; the high and moderate group are 1 - 3 points above this average, and the low GB group is well below average. Again, it wouldn't be surprising if higher groundball rates tend to be associated with a higher Batting Average on Balls in Play (BABIP). Groundballs are harder to catch than flyballs. But flyballs do more damage if you don't catch them. Therefore, groundballs produce a higher batting average than fly balls---but groundballs usually produce fewer extra base hits. BABIP is slightly correlated (r=0.12) with GB%, according to Steve Staude's correlation tool at fangraphs. Keep in mind that other components of BABIP, like line drives, may be more important contributors to BABIP--thus the relatively low correlation between GB% and BABIP.

LOB%: From the table, it would appear that Left on Base percent is higher for lower GB% groups. The differences are relatively small, leading me to doubt the significance of the variation among group's LOB%. The correlation tool indicates that LOB% and GB% are not correlated.

HR/Fly%: The league average HR/Flyball rate for the period is 10.8%. The high group is above average and the moderate and low groups are below average. The table shows HR/Fly rate increasing progressively at higher groundball rates. Some people theorize that GB pitchers will have a higher HR/fly rate because a disproportionate number of their flyballs are mistakes (e.g., sinkers that don't sink or hanging breaking pitches). However, this really hasn't been proven. Falling back on Staude's correlation tool, minimal correlation exists between GB% and HR/Fly% for his sample. And studies at Hardball Times and Fangraphs have reached varying conclusions, ranging from GB% produce lower HR/Fly rates to "little or no correlation." On the other side, Fangraphs' sabermetrics library states that groundball pitchers tend to produce higher HR/Fly rates, which would support the differences in the table. I attempted to analyze the variances for the three groups in the table, and the variation between the high and low group is statistically significant, but the same significance does not exist between the moderate group and either the high or low group. For my sample, there appears to be a difference in HR/flyball rate between the highest and lowest groundball pitchers, but any overall conclusion is tenuous.

ERA, FIP, x-FIP: The moderate GB% group has the best average ERA, followed by the high GB% group. The low GB% group has the worst average ERA. Both the high and moderate GB% groups produced an average FIP below the ERA, which is consistent with the overall average trend for MLB pitchers. The low GB% group has an average ERA below the FIP; this probably is due to the lower BABIP for these flyball pitchers (which is not reflected in FIP). The correlation tool indicates that groundball rate is slightly correlated with lower ERA (r= -.11).

The table below shows that the higher groundball groups had a lower average Line Drive rate.

LINE DRIVE RATE
HIGH	20.01%
MODERATE	20.75%
LOW	21.80%
MLB AVG.	21.20%

This is consistent with the correlation tool which shows that groundball rate is negatively correlated with line drive rate (r= -.22). The groundballs appear to be displacing some of the line drives that would otherwise be given up by the starting pitcher. This conclusion is not consistent with the usual view that groundball pitchers are expected to surrender a higher rate of line drives. In addition, at the individual level, line drive rates are volatile from year-to-year.

Cosart, Feldman, and Double Plays

The Beyond the Boxscore article demonstrated that relievers with elite groundball rates produce significantly more GDPs per opportunity, which in turn produces greater run suppression than pitching metrics like FIP would indicate. One would expect this is also true for starting pitchers with elite groundball rates.

I will examine the impact of GDP results for Cosart and Feldman, who posted GB% rates of 55% and 50%, respectively, last year. Both Cosart and Feldman produced ERAs significantly below their FIP. Because FIP does not account for above average GDP rates, the GDP rates may explain part of the difference between ERA and FIP.

I manually computed GDP rates per opportunity. GDP opportunities are defined as the following base states: first, first and second, first and third, loaded; zero outs or one out.

The American League average GDP rate is 10.9%. In addition to computing Cosart's and Feldman's GDP rate, I also computed a GDP rate for the six highest groundball rate pitchers from the "High" group in my table. This small group is intended to illustrate the GDP rate for the most extreme GB% starting pitchers. For older pitchers, I used the career GDP rate, and for younger pitchers, I used their 2013 GDP rate. In addition, I have shown the BABIP on groundballs for comparison to the ML average of .240.

GDP RATE PER OPPORTUNITY
	GDP%	GB-BABIP
Lowe	14.50%	0.208
Westbrook	16.10%	0.224
Morton	15.50%	0.238
Cahill	12.40%	0.229
Hudson	14.60%	0.211
Masterson	12.69%	0.213
Average	14.30%	0.22
Cosart	21.40%	0.175
Feldman	15.20%	0.164

A Scott Feldman GDP GIF is shown below

Cosart and Feldman both induced GDP rates much higher than league average. Cosart's 21% almost doubles the league average rate (10.9%). Feldman's GDP rate is almost 50% higher. The average GDP rate for the six extreme groundball starters is 14.30%, which is also significantly higher than the league average GDP rate. The GDP rate calculated for extreme GB% relievers in the BtB article is 19.9%, with individual rates as high as 29%. Although Cosart's GDP rate appears as an outlier compared to the starting pitchers, it is well within the range of GDP rates produced by elite GB% relievers.

I next estimated the run values for Cosart's and Feldman's GDP rates. I used -.85 runs as the run savings for a GDP, which I deducted from an average run expectancy of 0.9 for the associated base states. This allowed me to calculate the pitchers' runs saved aboved the league GDP rate, which I used to "adjust" Cosart's and Feldman's FIP downward. The downward adjustment for Feldman is 0.42, and 1.53 for Cosart. (The run savings are roughly similar to the RE24 for Feldman and Cosart.) The GDP adjusted FIPs are compared to the actual FIP below. (I am not suggesting that FIP should be adjusted for GDP percentage---this is a device for comparing GDP impacts to the gap between FIP and ERA.)

Actual FIP / "Adjusted" FIP

Cosart 4.35 / 2.82

Feldman 4.03 / 3.61

The adjustment erases all of the difference between Feldman's 4.03 FIP and his 3.86 ERA. The adjustment erases 64% of the difference between Cosart's FIP and his 1.95 ERA. My conclusion is that the GB rates, and resulting GDP rates, for Cosart and Feldman go a long way in explaining the apparent ERA over-performance. This may or may not be comforting---it depends on whether the GDP rates are repeatable. Feldman's GDP rate is close to the average for the six elite GB% starters, and seems sustainable. One could reasonably argue that Cosart's GDP rate is likely to regress to the 14% - 15% range. On the other hand, Cosart's is within the range of GDP results sustained by the groundball relief pitchers. It's possible that he can produce comparable results. The fact is that Cosart hasn't pitched enough big league innings to give us confidence as to what he can sustain.

Fangraphs notes that groundball pitchers tend to induce weaker contact on their groundballs, resulting in a lower groundball BABIP. As stated in Fangraphs' sabermetric library:

In other words, if a pitcher has an extreme groundball rate, they also tend to be extra good at making sure those grounders are hit weakly and turn into outs.

The GDP induced by Cosart, above, against the Rays is an example of weak contact. The BABIP for groundballs shown in the previous table for Cosart, Feldman, and the six extreme GB% pitchers are consistent with the finding that elite groundball pitchers tend to produce weak contact. Both Feldman and Cosart allowed GB-BABIP 65 - 75 points below league average. Each of the other six GB pitchers exhibts a GB-BABIP below the league average, with the average for those six pitchers 20 points below the league average GB-BABIP. Groundball pitchers produce more GDPs in part because of their groundball tendency. However, it's possible that these pitchers' groundballs are easier to field than the typical groundball---thereby leading to higher GDP rates.

An additional facet affecting pitcher's GDP rates is the defense's ability to turn double plays. Although the Astros' overall defense was bad, the Astros' middle infield was very good at turning the double play last year. The Astros were among the top MLB teams in GDP range, according to both UZR and DRS. Jose Altuve, alone, was +4 in GDP range (DRS). During the season, Bo Porter also credited the Astros' shifts for improving the team's double play proficiency.

Now we await the Astros' ground game in 2014.

Note: Fangraphs and Baseball-Reference were the sources for data in this article.