We all recall the frenzied internet hullabaloo -- an underrated word that should be used in daily conversation more often -- when the Astros traded minor league pitching depth success story Nick Tropeano to the Angels in return for catcher Hank Conger. In light of a bat that has not lived up to its PCL promise and the somewhat disappointing 78% stolen base success rate by opposing runners, the Astros faithful speculated aloud about the particular manifestation of insanity within the Astros' brain trust that led to such an incomprehensible swap.
Quickly, commenters and TCB writers were johnnies-on-the-spot with 30,000 words extolling the virtues of pitch framing, and how Conger stood to add three or more wins to the Astros by way of his ball receiving skills alone. The Great Internet Conger Fest of Christmahanukwanzaakah 2014 had commenced.
CSAA - A Mixed Approach to Measuring Catcher Framing
To this point, statistics that present run-prevention valuation to catcher framing skills were based on a fairly typical fixed-effects linear regression model that used ball and strike locations from Pitch F/X. The theory and math behind such numbers as StatCorner's Runs Above Average (for framing) were sound, but the good folks over at Baseball Prospectus -- namely Jonathan Judge, Harry Pavlidis, and Dan Brooks -- wondered if there wasn't something missing. They went to work, and on Thursday, published a piece describing how one can used a mixed linear regression model to account for random variation that had been otherwise ignored in the past.
If that seems obtuse to you, you are not alone...I had to look all this stuff up before trying to condense it into something Astros-centric and digestible. However, the BPro article is not behind a paywall, and I encourage you to read it - it sums up the differences between the two types of models extremely well in an easy-to-understand format. The "Gory Math" part is actually pretty easy to follow. Please read it.
The gist of the change is that the aforementioned fixed-effects models did not account for random variability represented by who the pitcher is, who the batter is, who the umpire is, and other inputs. But mixed models present a way to approximate the affects of these factors and incorporate them into the stat. The really cool thing is that since these random variations have nothing to do with Pitch F/X inputs, the approximation can be done using pitch count data available via Retrosheet all the way back to 1988 - there is more historical context than with other framing stats.
The result of this new method of modeling pitch framing contribution is a stat they call CSAA, or "Called Strikes Above Average".
Why You Should Care
The reason this matters to Astros fans should be obvious if you even grazed through the Baseball Prospectus article. Sitting comfortably atop the "present-day" CSAA leaderboard is newly-acquired backstop Conger. By approximating the effects of random variation that BPro's old framing model ignored, CSAA adds an additional five runs per 7,000 pitches beyond what had already been calculated before. (To put that into context, Jason Castro caught 8,655 pitches last season as a primary backstop.) For a full-time catcher, that means Conger would be worth almost one win more than the already-heartening older stat claimed he would be. At 37 runs per 7,000 pitches, Conger's glove could add between three to five wins to a club compared to an average pitch-framer!
To say that addition is significant would seem like hilarious understatement. Factoring in his offense and converting to WAR (fuzzy math involved, but that's okay, I'm only making a point about how much this means to the Astros), Conger's WAR could conceivably be equal to or better than players like Giancarlo Stanton, Adrian Beltre, and Andrew McCutchen. Conger is approaching "Mike Trout" level of value, in terms of Wins Above Replacement. As absurd as that sounds, the numbers -- all rooted in sound statistical methodology -- back up the statement. As an aside, is Buster Posey the most valuable player in baseball? He clocked in at 30 runs/7K pitches on BPro's chart. Add those 3-ish Wins to his already silly 5.7 WAR and you're approaching Babe Ruth territory.
Finally! Ausmus explained!
For eons, Astros fans endured the anemic bat of Brad Ausmus because we were told how wonderful a defensive catcher he was. Certainly, we saw few defensive miscues, and he did a pretty fair job at handling baserunners. But we were always told pitchers loved him. We love throwing to him, said Roger Clemens, Andy Pettitte, and Roy Oswalt. They never explained further, and so we were left to wonder why.
Baseball Prospectus' new model, because it's in part derived from an approximation of those random variables that are outside of Pitch F/X, can now be used to take a general swipe at catchers between 1988 and 2008 in order to figure out their pitch framing ability. Even with a level of uncertainty due to not knowing where in the strike zone pitches were thrown during that time, BPro still managed an impressive correlation level with their new stat. So without further ado, there are finally reliable-ish numbers with which we can make this convincing statement:
Brad Ausmus was, by far, the best defensive catcher of the last thirty years.
During his career, Ausmus saved an astounding 242 runs through his receiving skills, not including those saved by eliminating base runners. This pitch framing value adds almost 30 wins to his WAR total, bringing him to almost 50 for his career, putting him on-par with the likes of Cesar Cedeno, Mike Cameron, Jim Rice, and Jose Cruz.
The Lost Sheep
After a couple years of wondering whether Astros farmhand Rene Garcia would be lost in the Rule 5 draft, and after a disappointing season with the bat, we now learn a little more about the 25-year-old catcher who is now a member of the Phillies organization.
According to CSAA, which can be applied to the upper levels of the minor leagues due to the recent proliferation of Pitch F/X in AA and AAA parks as well as the new approximation of random variables, we now see that Garcia ranked as a Top-5 pitch framer at Double A in 2014, at 17 CSAA runs above average.
To call CSAA a revolution in understanding the defensive contributions of catchers would be unfair to those whose work preceded it. However, due to the rigorous thought applied to quantifying as many variables as possible, or at least estimating said variables, CSAA certainly seems to be the most complete method of defining a catcher's defensive worth.
People have been arguing catcher defense for over a hundred years, and we are still inching towards a more complete understanding. But beneath the noses of Astros fans, we have and have had two of the very best. My hope is that with understanding of how much value this skill adds to the player, that TCB readers will watch for Conger's effect on the game in 2015 - where he sets up to receive a pitch, how he brings the pitch into his body, how he reacts to pitcher location mistakes - and comment on whether or not they can "see" this in-game skill, where they might not have noticed it before.
As mentioned by the authors of the Baseball Prospectus piece, the CSAA model is very much a work-in-progress. The changing and out-dating of baseball statistics does not call their validity into question, as some doubters seem to suggest. Rather, it represents progress in understanding and defining the game we enjoy, and each iteration is a step closer to the whole truth.