Using Called Strikes to Find Sleeper Prospects: Catchers

While he’s not a sleeper, Will Smith’s (right) minor league performance has made him one of the best catching prospects in baseball. (via Dustin Nosler)

If you’ve been wondering whether Sicnarf Loopstok is a legit prospect, and not just an 80 grade name, you’ll be relieved to know he was the best pitch framer Double-A has ever seen, at least according to data dating back to 2008.

If you’re old enough to remember when Josmil Pinto was a legit catching prospect, we’ll explore why that was doomed to failure from the start.

If you’re a Mets fan, you’ve probably had countless sleepless nights wondering why Jeff Glenn was still with the organization, as recently as last season, despite his .191 batting average and 71 WRC+ as a 23-year-old at High-A. Spoiler alert, he’s a pretty good pitch framer.

If you were baffled when the Rangers called up Jose Trevino from Double-A, rest assured it was due to his long track record as an elite receiver.

Today, we’re going to delve into the metric of called strikes gained, otherwise known as called strikes above average, framing, or presenting. We’ll simply refer to this metric as CS%+ (called strike percentage plus), which will use a 100 scale to compare relative performance, with numbers greater than 100 indicating an above-average skill, and numbers below 100 representing below-average skill. Specifically, CS%+ is a measure of all pitches the batter does not swing at.


Attempting to measure something with too much precision can make a model less robust than a theoretically simpler, broader approach. While this may not be true for models where the inputs themselves are all pure, in cases where every single variable is an estimate, it may not be appropriate to assume greater precision than is available.

Note that this is not an academic assertion based on a deep understanding of statistics; rather, it is a pragmatic approach that prefers to acknowledge that if we have n factors feeding into a model, all of which are extremely significant to the model, one could argue the simplest approach is just to take each factor at an equal weight. The alternative would be to apply advanced models that strip out collinearity from these variables, then hope the new weights eke out a little more precision than just the simple average.

Where are we going with this? The probability of getting a called strike depends on a variety of factors, all of which are linked to each other to varying degrees depending on the league. The primary factors are as follows:

  • Catcher
  • Umpire
  • Pitcher
  • Batter
  • Venue
  • Day vs. night (7 p.m. starts get +0.3 percentage points more called strikes than 1 p.m. starts in Triple-A and the major leagues.)
  • Away vs. home (Away batters get 0.5-0.6 percent points more called strikes than home batters in Triple-A and the majors)

For the purposes of today’s piece, we’ll ignore the day/night and home/away effects, and focus on the other five factors.

Here’s a somewhat illustrative visual showing how much of a game’s called-strike percentage we can predict based on the first five factors, all the way down to Low-A, using all games played since 2008 that had at least 200 pitches thrown.

A few statistical notes:

  • The venue has a much stronger correlation the lower you go in the minors. This is likely due to increased collinearity with umpires and players due to the smaller league sizes. The R2 correlations are roughly 11 percent in Single-A, 10 percent in Double-A, nine percent in Triple-A and seven percent in the majors.
  • Umpires showed a consistent correlation from Double-A to the majors (8-10 percent range). High-A and Low-A were lower, but this may be due to flaws in the ability to accurately extract the umpire names from the box scores. Umpire names are manually entered and manually scrubbed. Strangely, Low-A showed a very strong 15 percent R2 correlation. Anecdotally, this author finds the data at Low-A to be superior to the High-A data for reasons unknown.
  • At the lower levels, pitcher ability here is paramount. Going from Low-A to the Triple-A, we get correlations of 30 percent, 20 percent, 21 percent, 20 percent, and 18 percent, compared to 12.5 percent at the major league level. Pitchers are consistently the strongest factor.
  • Catchers show a roughly 10-12 percent correlation in the minors and an eight percent correlation in the majors.
  • Batters show a similar 10-12 percent correlation in the minors and a seven percent correlation in the majors.

Now, one could pipe all these factors into separate models for each level and calculate the relative impact of each factor to attempt to capture the nuances of each level and factor. We will be taking a much simpler approach: For every pitch a catcher takes, we will calculate the average CS% (called strike %) of the batter, pitcher, umpire and venue. We will then divide his average CS% with his “expected” CS% based on the other four factors and voila, we have CS%+.

Validating the Model

Top MLB Catchers – 2018
Minimum 2,000 pitches received in 2018

How do we validate the model is functioning properly? We’ll compare it to another published MLB list and see whether the models are in general agreement, then examine where we have disagreement. Note that the model today is built with the express intent of building a model blind to pitch location since we don’t have access to minor league pitch locations. Thus, it is likely models that are location-aware will be more closely align with true talent.

First up, CSAA (called strikes above average), via Baseball Prospectus.

Baseball Prospectus CSAA compared to CS%+ | R2 0.25

The models, generally speaking, agree on who are the best catchers and who are the worst. Isiah Kiner-Falefa, Omar Narvaez, Welington Castillo, and A.J. Ellis were very poor according to both models. Similarly, both models agree that Austin Barnes, Roberto Perez, Jorge Alfaro and Sandy Leon are excellent receivers. Gary Sanchez and Mike Zunino are quality pitch framers. CS%+ really didn’t like Castro’s (small sample size) 2018 season, and I’m inclined to say that’s probably wrong since it liked him from 2014 through 2017 (as does BP).

Quick Tangent on Pitch Locations as Part of the Model

Leveraging pitch location will undoubtedly get you a better sense of catcher framing talent, as we can more easily ascribe a probability of whether a pitch was a strike. However, this also ignores game calling, since a catcher may want his pitcher to throw over the plate more or drop a curveball in for an easy called strike. While this doesn’t measure framing per se, it does capture the essence of one of a catcher’s responsibilities: maximizing their pitchers’ ability to get called strikes, be it by asking for a pitch on the edge or by grooving it down the middle if the situation warrants it.

What this implies is that a Prospectus-style model will not penalize a catcher if the ball is way outside (as I understand the model). CS%+ will naively assume the catcher should be able to influence the pitcher to throw the ball close to the plate and will thus capture some of a catcher’s ability to “calm a pitcher down” or “work with the pitcher” and similar intangibles that are difficult to measure.

Which Minor League Levels are Predictive?

How far down the minor league ladder can we go and still find useful signal in framing talent? Let’s take a look.

MLB CS%+ and Triple-A CS%+ | R2 0.36

We mentioned Josmil Pinto at the top of article. Clearly, he was a very poor receiver long before he ever made it to the majors. Had this author been armed with this knowledge back in 2014, he likely would have avoided drafting Pinto in fantasy with the hope of him staying at catcher. Ryan Lavarnway put up incredible offensive numbers for a catcher, with wRC+ numbers of 154, 145, 142, 137, and 172 between 2009 and 2011 as a young catcher in Red Sox system. His 91 CS%+ in Double-A, followed by his 94 CS%+ in Triple-A would explain why the Red Sox only gave him 250 plate appearances over three seasons between 2012 and 2014, despite the 34 home runs he hit in 2011. Lavarnway should have been transitioned to a corner position much earlier in his career.

Tomas Telis’ story doesn’t portend particularly well for Danny Jansen. Telis, much like Lavarnway, put up very strong numbers for the Rangers but quickly showed he couldn’t handle major league pitching. He posted a 95 CS%+ in High-A, 99 in Double-A followed by 98 in Triple-A before he reached the majors. Baseball Prospectus has a different opinion of Jansen’s framing success in the majors; however, the two methods are in agreement that his Triple-A framing was below average. Based on this catching metric, it would indicate Reese McGuire is more likely to stick as the Blue Jays’ starting catcher than Jansen. McGuire has been an excellent framer throughout his minor league career.

Our model was very high on Leon and Barnes, as well as d’Arnaud, Zunino, and Derek Norris. Generally speaking, it’s very difficult for catchers to improve a great deal over their minor league numbers. All of our elite major league framers were elite framers in Triple-A. Curt Casali is the only below-average Triple-A framer who has been able to crack 101+ in the majors. If you are not at least well above average in Triple-A, it doesn’t bode well for your future as a major league catcher.

MLB CS%+ and Double-A CS%+ | R2 0.18

We see a weaker relationship compared to Triple-A. However, Tomas Nido, d’Arnaud, Hedges, and Gallagher all were showing good receiving at this level. Carson Blair had a similar arc to the aforementioned Lavarnway and Telis–strong batting numbers–but never got any real playing time because he just couldn’t catch.

Double-A Catchers to keep an Eye On

  • Sicnarf Loopstok posted a 125 CS%+ last season, which was significantly better than any other catcher at that level. He also posted a very strong 113 CS%+ in High-A the season before. If he can translate his power to the majors, he has a good chance to stick and be something like the good version of Mike Zunino.
  • Tomas Nido isn’t a very good hitter, but his receiving skills should give him enough time to see if he can develop enough of a bat to stick.
  • Will Smith is legit. His 141 wRC+ combined with his 107 CS%+ means he has what it takes to be a major league receiver. Fortunately, Chavez Ravine is only a 29 min drive from Bel Air, which will make the commute easy.
  • Taylor Gushue, who was a little old for Double-A, posted a 108 CS%+ and likely will get a shot with the Nationals at some point.
  • Keibert Ruiz, who was really young for a catcher at Double-A, was only able to post an average CS%+. He was slightly above average in High-A and Low-A, so there’s a chance he can improve. If he can do so even a little bit, he has All-Star potential.
  • Reese McGuire recently graduated to the majors and played a bunch at Triple-A last year. As discussed above, he’s much more likely to stick at catcher than Danny Jansen.
  • Jose Trevino, mentioned at the top of the page, has been a fantastic receiver as far back as Low-A, posting 115, 112, 105, and 107 CS%+ numbers as he has progressed through the minors. He should be able to handle catching duties at the major league level despite the clearly sub-standard bat.

MLB CS%+ and High-A CS%+ | R2 0.08

We see our link rapidly deteriorating as we go down the minor league ladder. We also see some big misfires, specifically Sandy Leon and Austin Hedges. On the positive side, we were able to flag Danny Jansen and Tomas Telis as poor receivers as far back as High-A; Austin Barnes, Tomas Nido, and Gary Sanchez all had a good track record going back to High-A.

More Catchers to Keep an Eye On

  • Sean Murphy is an excellent receiver and posted 110 at High-A, followed by 103 at Double-A and 105 at Triple-A (very small sample size). Combined with his back-to-back seasons of 130 wRC+, don’t be surprised if he’s the starting catcher for the Athletics sooner rather than later.
  • Chadwick Tromp doesn’t quite have the ring of Sicnarf Loopstok. However, he was solidly above average as a receiver in Low- and High-A, but he has regressed slightly, so there may not a huge amount of potential here.
  • Ben Rortvedt, also of the awesome name club, posted a strong 106 CS%+ as a young 21-year-old with decent batting stats.
  • Jason Lopez has clocked in at a CS%+ of 104 in each of his last three seasons. Although currently blocked by Gary Sanchez, Lopez is only 20, giving him lots of time to develop. Given the strong base of receiving, youth and a 123 wRC+ at Single-A, Lopez is definitely a prospect to watch.

References and Resources

Eli Ben-Porat is a Senior Manager of Reporting & Analytics for Rogers Communications. The views and opinions expressed herein are his own. He builds data visualizations in Tableau, and builds baseball data in Rust. Follow him on Twitter @EliBenPorat, however you may be subjected to (polite) Canadian politics.
Newest Most Voted
Inline Feedbacks
View all comments
5 years ago

Tigers prospect Jake Rogers is said to be a talented defensive catcher in all defensive metrics.
So surprised as to his omission.

Yehoshua Friedman
5 years ago

Sicnarf is Francis spelled backwards. Does being spelled backwards help with pitch-framing?

5 years ago

How does the top catching prospect for the Cardinals, Andrew Knizer grade out under this system?

5 years ago

This is awesome research. Also, this is the first time I’ve read an article by you. Both of these pieces were news to me:
* Day vs. night (7 p.m. starts get +0.3 percentage points more called strikes than 1 p.m. starts in Triple-A and the major leagues.)
* Away vs. home (Away batters get 0.5-0.6 percent points more called strikes than home batters in Triple-A and the majors)

Both very interesting IMO, as the home field advantage in NBA and NFL is more pronounced than in MLB.

5 years ago

How did Austin Allen do?

5 years ago

ASTROS Chuckie Robinson? How did he pan out?