The new toy from Statcast, Catch Probability, probably isn’t ready for rigorous analysis just yet. It gives an estimate for the probability of a given batted ball being caught, but it doesn’t work for the infield, doesn’t pick up all batted balls, doesn’t account for different park dimensions, and is based on equations that those of us in the public can’t examine and evaluate. Some people think the metric presages a Statcast-based WAR, and while it might, it’s certainly not in the near future.
But despite all that, the information the metric provides is exciting. While it’s too raw for us to use to decide whether a given player is good, we can certainly use it to contextualize the evaluations we already have. It’s not going to settle whether Kevin Pillar or Kevin Kiermaier is the better center fielder, but it can shape our understandings of those players at the fringes, and show not just that they have value in the field but how they go about accumulating it.
It’s that last point that I want to turn to today. While the presentation of the data will surely change as it develops, right now individual batted balls can’t be matched with their catch probability. Instead, you can observe them sorted granularly by hang time and distance, or view them sorted into five broad buckets, which MLBAM has been referring to as “five-star plays,” “four-star plays,” etc. This method of presentation should be familiar if you’ve ever looked at Inside Edge data, which tries to do a similar thing with human stringers observing and classifying every batted ball.
What this presentation makes very easy is understanding how different people achieve their defensive values. If one player makes more impossible plays (i.e., the five-star plays) than anybody else, and one player makes every single routine play (i.e., the one-star plays) without ever making a mistake, they might grade out similarly in a broad measure of defensive value like UZR or DRS, but they do so with very different skillsets.
To try to get at this, I combined the data from 2015 and 2016 and recalculated every player’s rate of conversion in a given bucket from a percentage to a z-score. For example, Lorenzo Cain’s 21 percent conversion rate in the five-star bucket is above-average, and thus it gets re-expressed as a z-score of 1.3. Then, to distinguish between the “flashy” players (who get their value from converting more difficult plays than expected) and the “steady” players (who get their value from flubbing fewer easy plays than expected), I looked at the difference between their z-scores in the five- and four-star buckets and the z-scores in the other buckets, and the difference between their z-scores in the one- and two-star buckets and the z-scores in the other buckets*. The result is two numbers, a “flash” score and a “steadiness” score, that hopefully tell us something about how these players go about their defensive work.
*The extreme differences (the five- and one-star buckets, respectively) were weighted to be three times as heavy as the less-extreme differences (the four- and two-star buckets).
So who are the archetypal outfielders of each type, with the highest scores? I limited this inquiry to those players who had at least 100 recorded opportunities in 2015 and 2016 combined.
The flashiest fielders of 2015–16
Oh good, an excuse for some absurd Billy Hamilton highlights.
One more? One more.
Billy Hamilton is an incredible fielder, who makes plays that few other players can, as amply demonstrated above. His 44 percent conversion rate in the five-star bucket is the highest of 2016 among players with at least 20 opportunities. (It’s also the highest of players with at least 15, 10, or 5 opportunities, to give you a sense for how great that really is.) His high flash score is mostly driven by that massive differential; he’s fine at the other buckets, but none of them can compare to his performance in the most difficult one.
It’s an interesting mix of names that follow him. Kiermaier also shows up, and he — along with Jarrod Dyson and Travis Jankowski — follows a similar pattern, with above-average scores across the board that are most pronounced on the five-star end. Coco Crisp, on the other hand, sees his only positive scores come in the four- and five-star buckets, suggesting that aging appears to have most impacted his ability to make the routine plays.
Mark Trumbo is probably the weirdest name to see on here, if only because he’s more often associated with things like this…
…than with any kind of stellar defensive play. And indeed, Trumbo grades out poorly in every bucket, just less poorly in the more difficult ones. This makes some sense, though; it’s hard to be that much below average on extremely difficult balls to catch, because they are (by definition) extremely difficult to catch. There’s a lot more room to be awful on the routine plays.
What about the other side of the spectrum?
The most reliable fielders of 2015–16
The first thing to notice about this group is the narrower range of values. This makes sense, for the same reason that it made sense for Trumbo to show up in the prior list: you can only be so much better than average on the easy plays, and so much worse than average on the hard ones. That might also be why there are fewer truly great fielders on this list than the prior one; it’s hard to accumulate lots of defensive value through reliability alone.
Nori Aoki is also not the first player who comes to mind when I hear the word “reliable.” He’s certainly not known for his direct routes to balls:
But in a way, it makes sense that he’d show up here. Aoki is clearly athletic, and so even when something goes wrong (as it did above), he can make up for it, so long as the batted ball wasn’t a particularly difficult one to catch. Some of the players on the list, like Dexter Fowler, Denard Span, and Kirk Nieuwenhuis, seem to fit a similar profile, with good physical tools that are perhaps limited by poor jumps and reads. There are also some players who don’t have those good tools — Marlon Byrd, Seth Smith — which means what’s being picked up might instead be steady hands that lead to better-than-expected performance on relatively easy plays.
Again, the Statcast data is raw. But that’s a reason to be cautious about drawing conclusions with it, not a reason to not play with it. While it might not be ready to fundamentally change the way we view any given player, it can certainly help us better understand what we already think we know.