Looking At Robbie Cano And Defensive Metrics

Yesterday, Dave Pinto of Baseball Musings posted the 2006-2010 leaderboards at second base for his PMR defensive metric, which uses play-by-play data and factors such as batted ball type and how hard a ball is hit to measure defensive performance. The results are mildly surprising:

This is just the top 10, and I encourage you to visit Dave’s site and check out the remainder of the post, which breaks down Robbie’s performance further and has some interesting team defense data.

The reason that called these results mildly surprising is that based on a number of other metrics, Cano is not near the top of the second baseman rankings over the last 5 seasons. Here are his rankings for the last 5 years in some of the major available defensive metrics, with his “score” in each in the parenthesis:

UZR: 20th (-15.3)
UZR/150: 16th (-3.1)
+/-: 9th (+20)
TZ: 8th (+13.5)

UZR seems to be the outlier in that it sees Robinson as a subpar defender, while the others view him as solidly above average. Regardless, I think Cano provides a useful prism through which we can view the uncertainties inherent in using defensive metrics. This is not a particularly small sample, as 5 seasons should be enough to give us a reasonable amount of confidence in the results. Yet we have metrics that see Cano as one of the best gloves in the league, and others that do not like him at all.

Those of you who have read my work before know that I am not suggesting that the metrics have no value or that using them to build arguments is a fatally flawed exercise. But it is important to know that they have limitations, as I have previously noted:

The defensive metrics that we have are far from perfect. They are all subject to sample size issues, some have inherent biases, and all must cope with the fact that with the technology currently in play, we are unlikely to measure defense with any semblance of exactitude. Additionally, some of the more popular metrics that are based on batted ball and hit location data compiled by video scorers, such as UZR and +/-, are beholden to evaluations that are at least partially subjective. For example, in an excellent Baseball Prospectus article yesterday, Colin Wyers discussed range bias, in which a players range influences his expected outs. Put simply, it suggests that a player with poor range can actually have his UZR or +/- inflated because the scorer who is marking the game will mark a ball that he could not reach out of the player’s zone incorrectly. This subjectivity means that these metrics should not be taken as gospel, and are not fit to be used to evaluate defense with the same degree of confidence as wOBA or FIP might be used to measure other elements of the sport.

So what do we do? How do we evaluate defense? At this point, the best method is to use all of the data that we have available to us, both statistical and observational. Look at UZR, DRS, TotalZone, Fan Scouting Reports, nFRAA, fielding percentage, errors, putouts, assists, other fan opinions, the thoughts of scouts, and your own observations as well, weighing each factor as you see fit. When the composite of those inputs allows you to reach a conclusion, try to avoid presenting that conclusion as being exact or perfect, as a different weighting of the inputs would likely yield a disparate result. In sum, use the metrics, but be prudent and always be cognizant of their limitations. If we treat them with greater respect and reverence than they deserve, then much like I said regarding projections last week, we are in danger of reaching conclusions that are not supported by the evidence.

Follow Me On Twitter