Seattle Sports Insider

State of Defensive Sabe

extra points for stride "extension" there, Mitch

Posted by

12/26/16

This Dec. 22 Hardball Times article is cool for its --- > depiction of the way that ML coaches and players are reacting to 2017 data.

Kevin Keirmayer, for example, draws swooooooons at BJOL for his defensive prowess; the numbers love him but "I don't love them back," says Keirmayer. No, wait. That was Pillar with that exact quote. All of the managers and players say the same thing, though: they know what UZR is, they follow it, are well aware that (fangraphs) WAR is driven by UZR. But wayyyy too often it contradicts what they know to be true. As Bill James said, "If your defensive system tells you Johnny Bench is a bad catcher, make another defensive system.

In fairness, my friend, it has been a long time since USSM was shouting down everybody who dissented from a 1-year UZR number. THESE days, I don't know of a single place on the 'net where UZR, itself, is taken as gospel. But! Neither is RBI gospel, and we want to know a player's RBI, right? We're back to using all stats with a sense of proportion.

A few quotes on UZR specifically:

“They just put me in the lineup and I play,” says Nationals left fielder Jayson Werth. “I do know that the UZR stat’s convoluted because it’s dependent on a person to collect the data. Then, you know, WAR’s affected by UZR as well – one version, depends on where you get it, I guess. So, I don’t put a lot of weight in UZR and I don’t think it’s that accurate of a stat.”

“I’ve seen the defensive statistics that say that Hechy (Adeiny Hechavarria) is not a good shortstop, and that’s just stupid,” says Marlins outfielder Christian Yelich. “You can’t say that and have any sort of credibility at the same time. I think a lot of sabermetrics and a lot of the numbers don’t tell the whole story. You’ve got to watch the game, as well. You can’t just look at a sheet of paper, look at what it says, and say, ‘This guy’s good, that guy’s not good,’ just based on looking at paper.”

“I get the info on that stuff, and I see some where they might rate so-and-so better than another guy and it shocks me, because we play these guys and I’ll just go, ‘There’s no way in the world that’s the case,’ in my mind,” Toronto manager John Gibbons says.

Another manager goes on to say, though:

“Here’s where I’m into the numbers – it’s for the guy that’s not so obvious, the guy that hasn’t made his mark yet,” says Cubs manager Joe Maddon. “If you can accumulate enough information about that particular guy, and you can tell me why he’s going to be good in advance of him being good, that’s where I like the number. But I already know who’s good. You know who’s good, he knows who’s good, we all know who’s good.

This makes sense for the SSI denizen also. We are not nearly as interested in Kyle Seager's defensive numbers, as we are Mitch Haniger's next year. What would we "do" about Kyle Seager anyway? It's a key concept.

Fascinating is the fact that the men in uniform are irritated by UZR's problems -- as we would be irritated by RBI's problems, if RBI were gospel -- but that they are quite fond of StatCast defensive numbers.

:: taps chin :: It might be fun to have a "StatCast Corner" during the season, unexpected measurements of M's players that move us in one direction or another.

Pillar sees the system grading players for the right things... “I think Statcast is awesome,” Pillar says. “It’s not a perfect science, but it gives you an idea of a guy’s ability to use instincts and read. Obviously, that’s judged by first-step quickness. Route efficiency is great to tell you how well guys run to the ball. I think those are useful statistics. … I think that’s some of the most important things for being a good outfielder – your ability to get a read, and your ability to run the route. Top speed, obviously, is important, but if I have a quicker first step and run a better route, I’m going to cover more ground than the guy who’s faster. So, those numbers are leading toward better defensive metrics.”

So, a second big takeaway. Keep an eye on first step and route efficiency, even on a single play in isolation; the attempt to reverse-engineer outs made can be inherently more problematic for all the moving parts contained.

Corollary: the teams have superb data on these kinds of numbers internally. We've reached an ebb point in the 100-year sine wave, where it is much harder for fans to argue with the teams' decisions on defense. If Dipoto thinks Haniger can or can't help the club in center, he's likely more right than we are.

BABVA,

Dr D

Blog:

Sabermetrics

Tags:

Kevin Keirmayer

Comments

We're getting to that point

We're getting to that point where we desperately need a better defensive metric publically available. The biggest problem with UZR is that it doesn't filter human bias. The whole point stat based analysis is to remove human bias.

I like statcast a lot, but its also difficult to credit fielders for good positioning. Its much more difficult to measure good fielding than it is for offense. You're relegated to scouting the player yourself or using educated guesses based on the numbers and scouting reports.

Funny Numbers on FanGraphs

The day of the trade, Jean Segura was rated a 2.7 WAR player on FanGraphs, Mitch Haniger at 2.3. Today they are rated at 2.1 and 1.7 respectively. What? Safeco costs over a full win on two players? Then Ketel Marte is today listed as a 0.0 defender (league average?), while Jean Segura, widely considered a superior fielder, is listed as a -4.7 defender. But then again, on Segura's player page, the depth charts listing is as a +1.9 defender. I think FanGraphs' calculators have a poor first step and even poorer routes.

We hear he is a whiz of a Wiz, if ever a Wiz there was

Blame Ozzie!

When some of us were kids and Belangers, Hansons and Aparacios were the standard of glovey SS's, you rarely saw the SS backhand one to his right, plant and then gun. "Fundamentals" dictated that you got both feet in front of it, half-skipped and then flipped it sidearm to 1B.

But then Ozzie came along. The Wizard changed everythig for SS's. It wasn't enough to make the plays, you had to look spectacular doing it. The play to the right was born. It didn't hurt that turf fields were becoming the norm, as well. Turf meant ball speed, so SS's could play deeper...had to play deeper. I think the spectacular play, rather than the two-handed scoop, is what lights defensive metrics now. UZR wants to see something that looks hard....not that looks easy.

Get this:

Belanger had a career Rf/9 of 4.93 (Single Season high of 5.47) and a career Fld. Pct. of .977

Aparacio 4.89 (5.7) .963

Hanson 5.03 (5.29) .962

Ozzie Smith 5.22 (5.87) 9.66 (his range was in a league of it's own)

Cal Ripkin 4.73 (5.50) .969

Segura 4.48 (4.73) .974

And last year's two GG winners:

Crawford 4.63 (4.81) .974

Lindor 4.20 (4.37) .975

GG SS's don't get to nearly the amount of balls as SS's of the past. Ripkin, for example, was a multi-GG winner, but never known for his range. He would get to 80 more balls a year than Lindor does. Amazing. The previous 4 GG's in the Al were won by Escobar and Hardy. They get to 4.26 and 4.33 balls per game (Rf/9).

K's are up, for sure. But I wonder how much of the decline in range factor is due to deeper positioning and the reluting back-handed play. In a good year, even Ripkin would get to 190 more balls than Hardy. How amazing is that?

Ozzie didn't invent the play to the right

And UZR isn't wanting to "see something that looks hard." Since baseball began, SSs have had to range to their right and to their left. Their positioning has evolved and their abilities have as well, but the play to the right always existed and always will. Ozzie Smith influenced a generation of SSs to emphasize style in their movements, but, ultimately, the balls go where they go and if you don't catch them, the league figures that out.

As for UZR...it's not looking for plays that "look" hard..it's looking for plays that are statistically more rare. Its leading assumptions are frequently wrong, it doesn't account for the team cooperation involved in defense, it doesn't account for positioning, it doesn't account for scorer biases, and it doesn't account for the fact that statistically rare plays are rare events and need to be treated with rare-event statistical methods and not linearly. It's a bad system, but it's the best bad system we had for a long time that was based on the fielder-in-isolation model. I preferred team-based results-oriented metrics like Win Shares and my own marginal defensive ratings because they were more stable and did account for positioning and team cooperation (and ignored the possible biases that come from the scorer, for the most part), but even those approaches are bad compared to actually tracking the plays and the fielders' physical abilities in real time.

My .02$