clock menu more-arrow no yes

Filed under:

Sabermetrics: Hard & Soft Contact

New, 25 comments

Astros Sabermetrics: How Should Pitcher Soft and Hard Rates Be Evaluated?

Jason O. Watson/Getty Images

When the Astros acquired Mike Fiers, some commenters expressed concern because Fiers had the highest percentage "hard contact" this season, according to Fangraphs' soft/medium/hard metric.  However, as I noted in my article prior to the trade, Fiers was one of the few potential pitcher trade targets who has been above average in "soft contact."  Obviously one would prefer a pitcher like Dallas Keuchel, who is No. 1 at avoiding hard contact and No 3 at inducing soft contact.  But that is an exceptional situation.  What should we make of a pitcher like Fiers who induces both soft and hard contact?

When I wrote the pre-trade article on buy low pitcher trade targets, I focused on the soft contact rate, but ignored the hard contact rate.  The hard contact rate is likely to be driven by the pitcher's HR/fly ball rate, which is likely to regress toward league average (according to sabermetric theory).  Moreover, the purpose of the article was to identify "buy low" pitching candidates, and a high HR/Fly ball rate helps us identify buy low pitchers, since they are likely candidates for a rebound.  Soft contact, on the other hand, is more likely to be a pitching skill.  Whether it is a repeatable skill may be up for debate.  But enough of the pitchers seem to be good at repeating soft contact management over their career that it appears to be an enduring skill for the elite contact managers.

Let's dig more deeply into soft and hard contact characteristics.

Hard Hit Balls and HR/Fly Rate

Fangraphs' glossary provides a succinct explanation for normalizing HR/Fly rate (based on league average) in the x-FIP formulation:

...we use xFIP to see how a pitcher might be expected to perform given an average HR/FB% because we do not expect pitchers to have much control over that number. They can control how many fly balls they allow, but only a limited set of pitchers can truly influence their HR/FB%.

If "hard" hits are driven mostly by HR balls and the objective is to predict a pitcher's future performance, then the pitcher's hard hit ranking may be less relevant than the soft contact rate.

Fangraphs' pitching correlation tool can be used to test whether HR / fly ball rate has predictive power over the next year HR/fly rate.  The base correlation between Year 1 and succeeding Year 2 HR/ fly ball rate is .08---essentially no meaningful correlation at all.  This supports the notion that pitcher HR/Fly rate can be expected to regress to league average. found that batters influence batted ball exit velocity five times more than the pitcher--which also is consistent with normalizing the HR/ Flly rate.

I performed correlation analyses for 92 major league starting pitchers in 2015.  Correlation between the soft and hard contact percentage and HR/ Flyball is shown below.

Correlation With HR / Fly Ball Percentage

Soft   -0.1419

Hard  0.343

Soft contact is related to HR/fly rate in the exxpected direction (negative), but the correlation is relatively small.  Hard contact has a stronger correlation with HR/ fly ball rate.  This seems to confirm my previous assumption that the hard contact classification is influenced to a significant degree by the pitcher's HR / fly ball rate.

Correlation With Other Characteristics

Based on 2015 pitcher data, I examined the correlation of soft and hard contact with other characteristics, in order to identify other possible factors influencing the measures.  The correlations with batted ball characteristics is shown below.

Correlations With Soft & Hard Contact
BABIP -0.080 0.170
Line Drives -0.208 0.225
GB% 0.324 -0.478
Infield Fly 0.170 -0.040

The most significant correlation showing up in this comparison is ground ball rate.  A higher ground ball rate is likely to induce more soft contact; hard contact is inversely correlated with ground ball rate.  Keuchel's elite ranking in both soft and hard contact is consistent with this result, given his extreme ground ball rate.  To some extent, this may explain Fiers' hard contact rate, since Fiers is a fly ball pitcher.  Colin McHugh has a moderate ground ball rate (45%) and exhibits a very good soft contact percentage, plus a somewhat below average hard contact percentage. Newly acquired Scott Kazmir's ground ball rate is approximately average; he has a below average soft contact rate and significantly less hard contact than average.

Hard contact also has a correlation with BABIP.  Similarly hard contact is correlated with line drive rate, while soft contact is inversely correlated with line drive percentage.  For pitchers, both BABIP and line drive rate are volatile measures which fluctuate over annual periods.  McHugh and Fiers currently have well above average BABIP (.314 and .315, respectively).  Considering McHugh's good contact results, presumably this increases the tendency for his BABIP to regress toward league average (.293).  In addition to normal regression, Fiers' BABIP may benefit in the future from the Astros' defense.  Fiers has moved from one of the worst defensive teams to one of the best.  Keuchel has a below average BABIP (.263).  Although this may signal some degree of regression, several factors may suppress his BABIP: excellent soft and hard contact percentages, the Astros' team defense, and Keuchel's own fielding ability.  Last year, my article on pitcher fielding showed that Keuchel gained a full win from his fielding, and suggested that the best fielding pitchers tend to be soft contact specialists.

Infield fly ball rate is correlated with soft contact percentage.  Fiers is above average in his rate of inducing infield flies, which may explain why he can produce above average soft contact despite a low ground ball rate.

I also examined the extent that pitch types are correlated with soft and hard contact.  Curve ball and change up usage does not show a significant correlation with hard or soft contact.  However, soft or hard contact frequency is correlated with 4 seam and 2 seam fastball usage, as shown below.

Correlation With Hard/Soft Contact

Soft Hard
4 Seam FB 0.045 0.284
2 Seam FB 0.12 -0.32

As indicated above, 4 seam fastball usage has little effect on soft contact frequency, but it is correlated with hard contact rate.  Although much has been made of McHugh's low rate of fastball usage, perhaps this tendency has enabled him to achieve a below average rate of hard contact.  Two seam fastball usage shows a moderate effect on soft contact frequency, and a more significant correlation with avoiding hard contact.   Fiers does not appear to use a two seam fastball, which may increase the liklihood of hard contact.

The usage of two breaking pitches, sliders and cutters, appear to be related to hard and soft contact frequency.

Correlation With Hard/Soft Contact

Soft Hard
Cutter 0.11 0.41
Slider 0.27 -0.16

Cutter usage shows an odd relationship.  More cut fastball usage shows a small to moderate correlation with soft contact frequency.  However, increased cutter usage is correlated with a higher hard contact rate.  It's unclear whether this is an anomaly or represents a real effect.  If cutters are used mostly against opposite handed batters, perhaps that explains the degree of hard contact.  Fiers relies upon a cutter more than most pitchers.  Increased slider usage is correlated with more frequent soft contact. and has a moderate inverse relationship with hard contact.  Since Fiers no longer appears to rely upon sliders, this may also affect the degree of hard contact.  Keuchel uses a two seam fastball, a cut fastball, and a slider, all of which are correlated with either avoiding hard contact or inducing soft contact.

Obviously there is more to contact management than just pitch selection.  Exit velocity data is a different metric than fangraphs' soft/medium/hard contact, but it also attempts to measure the quality of contact.'s evaluation of pitchers with the lowest average exit velocity indicates that control and command are  important factors:

These pitchers control their opposition’s quality of contact partly by driving the hitters into bad counts. In pitcher’s counts, hitters tend to put weak, defensive swings on the ball, resulting in glancing contact. About 15 percent of pitchers’ exit velocity suppression comes from controlling the count.

Exit Velocity Data for Pitchers

Exit velocity data publicly available from the new statcast metrics is both similar and different than the fangraphs' soft and hard contact frequencies.  The similarity is based on the objective of measuring the degree of hard or weak contact.  The difference is that the fangraphs data is based on the frequency of batted balls within soft, medium, and hard categories, while exit velocity is based on average miles per hour.  In addition, the fangraphs classifications are based upon determinations made by stringers or analysts, while calculation of exit velocity is automated by electronic devices installed in the ballparks this year. But keep in mind that the proprietary hit f/x type data available to teams may be more reliable than the statcast metrics.

The article ranks 486 pitchers by contact quality management based on the exit velocity data available at the time.  Keuchel, McHugh, and Feldman are ranked 5th, 10th, and 23rd, rankings which are nice to have in the same rotation.  Kazmir and Fiers are ranked worse than average on this metric..  However, based on recent information about the statcast exit velocity data, take these rankings with a big grain of salt.

The pitcher exit velocity data should be used and interpreted with caution.  This Beyond the Boxscore article lays out the potential flaws and inaccuracies in the statcast exit velocities currently made available to the public. In addition, Fangraphs' Tony Blengino identifies inconsistencies in the statcast exit velocity data in his article, "The Limitations of 2015 Statcast Data."

The problem derives from the fact that the statcast exit velocity data is incomplete. As Blengino points out, the magnitude of missing data is large, relative to previous hit f/x data. (Note that Minute Maid Park has the lowest percentage of missing data.) The statistical concern arises because the missing data is not randomly distributed.  In particular, soft contact ground balls and infield fly balls appear to be over represented in the missing data.  Blengino notes that this may understate the contact management skills of pitchers with high infield flyball rates, as well as extreme groundball pitchers like Keuchel.  Hopefully, continued refinement of the exit velocity data will result in more reliable data in the future.

Arguably, the fangraphs' soft/medium/hard metric may be a better resource than the exit velocity data in the near term.