Banner Report: The Linebackers
March 4, 2013Kyrie Irving probable for game against Knicks; Waiters, Zeller questionable
March 4, 2013I’m going to be honest with you: I have my misgivings about advanced defensive metrics—especially those that are publicly available to schmoes like you and me. 1 It’s not that I think the methodology behind these stats is flawed or that the approach itself is a silly one—it’s just that all these metrics seem a bit too reliant on two somewhat flimsy inputs.
First, there is the simple problem of sample size: it is unlikely that a single player will accrue enough defensive opportunities on different sorts of batted balls in a given year to give us a realistic impression of what his actual skills are. It’s been said that you need three years of defensive data—I assume playing almost every day—to make up for this paucity.
The second dubious input problem has to do with the data collection itself: the defensive metrics all rely on “stringers” who are actually at the game and tasked with classifying batted balls into categories. For example, was that ball that Shelley Duncan gracefully tripped for a liner or a hard hit fly ball? Maybe it was a fliner? Who knows? Which sector of left field was it hit to, section 32 or 33? Sometimes these stringers are allowed to record data only at their home ballpark, which can introduce all sorts of biases into their data collection. In short, this is a tricky business and particularly prone to human error.
Really though, it’s all we have. Even though we clearly don’t have the best data out there, we are light years ahead of where we were just ten years ago with respect to measuring individual player defense. The reason, of course, that I lay out such a long-winded introduction to the state of current defensive metrics is that I want to briefly consider some of them, especially in regard to our outfield makeover.
I believe it’s fairly well-accepted that our outfield defense should be improved this season. For one, we’ll have something resembling an everyday left fielder, which is a great start. For two, all three of our outfielders have spent considerable stretches of their careers (in some cases, their entire careers) playing center field, typically a sign of elite defensive talent. For three, there will be none of this, as much as we may well miss it.
All of those are sort of subjective measurements though. It’s great to think we’ll be improved defensively, but really, how many more games might we win because of that improvement? This is where we have to introduce those defensive metrics, warts and all.
So first, I tried to grab both total zone rating and UZR (two different ways to skin the same defensive cat) for the guys who played OF for us last season. I left a few guys out who played very sparingly (Jason Donald, Brent LilliPUTIANbridge), but for the most part, these are the guys who roamed the power alleys for us last year, along with two measurements of how many runs they saved or (more often) allowed compared against the average player.
Name | OF-UZR | R-TOT | AVG |
Aaron Cunningham | 1.5 | -3.0 | -0.8 |
Michael Brantley | -1.9 | -10.0 | -6.0 |
Ezequiel Carrera | -3.1 | 3.0 | -0.1 |
Vinny Rottino | 2.4 | 3.0 | 2.7 |
Johnny Damon | 1.1 | 1.0 | 1.1 |
Thomas Neal | -2.9 | 0.0 | -1.5 |
Russ Canzler | -1.8 | -1.0 | -1.4 |
Shelley Duncan | -3.1 | -2.0 | -2.6 |
Shin-Soo Choo | -17.0 | -15.0 | -16.0 |
TOTAL | -24.4 |
Some stalwarts may be surprised to see how terribly Choo and Brantley played. I’ve heard pundits occasionally laud both for their defense, but it’s become more and more accepted in recent years that Choo had only one strength (arm) with quite a few weaknesses (range, first step, going back on balls), while Brantley still profiles as a below average center fielder due to his average range and below average arm. It should be noted that he looks just dandy2 in left, as his weaknesses are less pronounced there and the average left fielder is much worse defensively than the average center fielder.
Anyway, these metrics suggest that our outfield allowed 20 to 30 more runs than the average MLB outfield would have over the course of the 2012 season.
Now let’s look at what our new outfield will look like. I had to make a few assumptions here. First, I looked at a player’s entire career and did my best to limit it to the particular position he’ll likely be playing in 2013 for the Indians. So for instance, I looked only at Brantley’s performance in LF over the course of his entire career. In Stubbs’ case, this wasn’t possible, as every inning he’s played defensively has been in center, despite that fact that he’ll be moving to right for the Indians. I’d therefore consider this a somewhat conservative estimate. Balancing that conservatism, we’re assuming that whoever fills in for these guys on their off-days will be average defensively, which is likely a stretch. Hopefully these two assumptions balance each other out. Anyway, here’s the data:
UZR-150 | R-TOT/yr | AVG | |
Drew Stubbs (OF) | 3.9 | 13 | 8.5 |
Michael Bourn (CF) | 10.7 | 10 | 10.4 |
Michael Brantley (LF) | 3.3 | 8 | 5.7 |
TOTAL | 18.8 |
Granted, this methodology isn’t perfect, but just by taking the average of the two most accepted publicly-available defensive metrics suggests that our new outfield could be a net improvement of around 45 runs, which would result in four to five more wins, strictly from improved defense at three positions.
There are all sorts of caveats here that I’m not going to get into. Michael Bourn could lose a step as he enters his 30s. Drew Stubbs may not adjust well to right field. The defensive metrics may be flat out wrong. Or injuries, MY GOD THE INJURIES!!!
But using the data we have available, it’s possible that without swinging a bat or stealing a base, our new outfield could win five games strictly by turning batted balls into outs at a more efficient clip than their collective predecessors. That’s at least something for our pitching staff to dream on.
AP Photo/David J. Phillip
- Craig just got back from the Sloan Analytics Conference, and I’m dying to talk to him about it. One item I’m hoping to discuss is the movement toward proprietary, team-housed analytics, especially in regard to defensive valuations in baseball. This is where it’s going, methinks. There was a decade or so where the schmoes were outsmarting the teams, but I’m pretty sure that’s over now. They’ve bought out the geniuses and their IP, and moved it behind closed doors. The “Moneyball Era” was exciting. But let’s face it: it’s over. [↩]
- Marte [↩]
22 Comments
Footnote 1 might be the nerdiest thing these nerd eyes have read in a long, long time – and I just finished reading A Game of Thrones. Wow.
But don’t I make up for it with footnote 2?
I’m not sure you understand how this works. You don’t overcome being a huge nerd by following-up nerdery with more nerdery. What, were you homeschooled?
“misgivings about advanced defensive metrics”
Amen.
that said, the OF defense should be improved. quantifying that to runs is a difficult to impossible task and I applaud you for making an effort in order to demonstrate the importance of our OF defensive improvements.
I wouldn’t call it difficult to impossible. Certainly not the latter. We have a pretty good idea of how many runs a team defense (i.e. not pitching) is allowing. Even if we don’t accept the metrics, I think it’s pretty much agreed upon that we are going from a below average performer at each position to an above average performer. Whether the difference is 30 runs, or 60 runs, we expect it to be quite a lot.
yes, so we can analyze it qualitatively rather easily (i.e. better by alot). analyzing it quantitatively is rather difficult (giving an exact number within a small margin of error). that is all.
If the problem is that we require a small margin of error, then sure. But quantifying runs saved is no impossible task, and the margin isn’t so wide that we should just throw our hands up and go “oh well”. And our projections at it have no less a margin for error as our projections for offense, which we still have no problem taking a run at.
En fuego!
Are you married?
At this point, I think I should note for the record that 3 of the 4 members of my immediate family either were or are now homeschooled.
I will not be mocked by a man with a Daenerys poster on his ceiling.
She is the dragon.
and now we all understand you 🙂
Do you? There’s a 75% chance, but as I always say, stats lie.
this understanding was a qualitative study, so yes.
I hope your concern with my personal life is well-intentioned. But I’m assuming that this is just a personal attack, which you chose because you have nothing useful to add to the discussion.
Oh Steve come on now so defensive and insulting to boot. I’ll mark down NO for your file. Have a nice day!
Really? Better by a lot?
Qualitative analysis in sports is far from easy or reliable.
If anything, the quantitative measures being unreliable just shows how much more suspect the qualitative is. I mean, if the numbers show that you need three seasons to get a useable sample size then thinking a guy “looks good” over a week or a month or even a year is obviously way suspect.
I would argue it is easy but not reliable. It’s easy to say that the Indians OF defense will be better by a lot because we are putting 3 players in place who have the ability to play CF (the most difficult of the positions and the one that requires the most range). That does not make it reliable (though the added quantitative quality of Jon’s writing helps add to it).
I completely agree that using the week, month, and oftentimes even year to evaluate a player is suspect. Not as if that will stop us from doing it in the future 🙂
Bringing it back full circle… nice!
Yes yes. Very much agree with this. It is one thing to say, “I am uncertain, but it is very likely to be between 20 and 60 runs better than last year.”
It is quite another to say, “It will be exactly 43.798963245 runs better.”
Obviously the second statment is silly. But I’d much rather say the first statement than to say “The defense will be better, but due to inexact metrics, we shouldn’t even fathom a guess as to how many more games we might win because of that improvement.” That seems to needlessly throw one’s hands up rather than accepting that ALL forecasts have uncertainty built into them.
If the defense will help the team win more (as seems obvious), then we should say why we believe it, and how much it might help. Whether our error bar is 2 wins or 20 wins, we should at least be taking stabs at trying to get away from nebulous and subjective statements about “improved defense” without ever demonstrating why improved defense might help a team win more games.