On Arms
by Jeff Sullivan on Jan 22, 2009 4:47 PM PST
in Seattle Mariners Stats
Email
PrintJust to add to what Matthew talked about below...
With a statistic like this, where you're looking at guys having ~between 100 and 200 opportunities a year, you'll want to look at several seasons' worth of data in order to get a more accurate picture of how good or bad they really are. One season just doesn't tell you very much about a guy's arm. It can give you an approximate idea of how well he performed at the time, but if you're trying to figure out a player's true talent level - which is usually what we're going for - then you need to expand your viewing window.
So with that in mind, let's look at a few interesting players:
- Ichiro had yet another season of being solidly above average, in both center and right. Since 2004, he has a 141 Kill+ and 117 Hold+ in CF, and a 118 Kill+ and 111 Hold+ in RF. I don't know if the inflated kill rate when playing in the middle is real, but those are impressive hold rates. People don't run on Ichiro. He has a reputation as being powerful and lethally accurate, and while the latter is a little bit of an exaggeration, he throws just enough runners out every year to deter many others from trying. This is just another of his many valuable skills.
- Franklin Gutierrez had a reasonable year last year, finishing just above the average in both categories. However, he was below average in limited trials in 2006 and 2007, so I wouldn't go too crazy just yet. The good news is that we didn't get him for his arm. We got him for his legs. Anything else he contributes in the field will just be a bonus.
- Adam Dunn hasn't posted a Kill+ above 75 since 2004, and he hasn't posted a Hold+ above 97 since 2005. Arm ratings aren't yet included in the UZR available on Fangraphs, so if you find yourself looking at Dunn's player page, you might as well subtract four or five more runs from his overall value in the field. He really ought to get over his pride and just donate his glove to charity or, failing that, a group of pioneers trapped in the Sierra Nevadas. They could put it to better use.
- Raul Ibanez has been posting good kill rates since the beginning of this statistic. He doesn't have the strongest arm in the world, but he has an accurate one. Not everything he does in the field is a disaster.
- Brian Giles has as much power left in his arm as he has left in his bat.
51 comments
|
0 recs |
Do you like this story?
Comments
Accuracy
Raul Ibanez has been posting good kill rates since the beginning of this statistic. He doesn’t have the strongest arm in the world, but he has an accurate one. Not everything he does in the field is a disaster
the other angels fan
There was a mosquito on that blade of grass
and Raul is adamant about wiping out malaria.
You’re welcome.
I love the YES HD title at the end
NEEDS MORE FREEDOM!
by Scruffy Lefty on Jan 23, 2009 7:59 AM PST up reply actions
I feel like Manny and Raul have high Kill+ is because no one respects their defense.
They both don’t have strong arms, but people will be more likely to try and get that extra base when they shouldn’t against those two.
That's probably because they're counting on Raul stepping on the ball causing his feet to fly out from under him as he careems off the field and into the ball girl in a Stoogean display of slapstick defense.
Or in the case of Manny, they figure he’ll just refuse to throw it back until he gets a 5 year $120 million contract extension.
So how would you rank the importance of an arm?
I was under the impression that it was like Range>>>>>>>>>>>>Arm>>hands/fundamentals/smoothness, etc.
This is the first year I’ve noticed John Walsh’s work on the subject. It sounds amazing to me to think that an outfielder can be worth a full win with his arm. Also…
- You say this is an inferential statistic like the fielding metrics. How would you rank it compared to the BIP metrics?
- Even if it is on par with them, wouldn’t we want more data points as a proxy in similar fashion as to checking UZR, PMR and the Fielding Bible and then finding a conclusion in between the three?
Range is the most important factor, and most arms are -5 < x < +5 runs
But arms still matter, and for some players a great deal.
I don’t know how I’d rank it compared to the BIP metrics. I like it, but I haven’t worked with it that much. I imagine you need more years’ worth of arm data than you do of defensive data in order to arrive at a reasonable conclusion, due to the limited sample sizes.
In theory you always want more data points, but where defensive metrics have some disagreement between their methodologies, I don’t see how another system could really have many disagreements with Walsh’s formula.
by Jeff Sullivan on Jan 22, 2009 11:06 PM PST up reply actions
There are some factors that affect the hold and kill numbers besides arms
For example, B.J. Upton is very fast, and plays a shallow center field on turf. So in the single with a runner on second scenario he’s going to have an edge from factors other than his arm. I have no idea how, or even if, you could really factor that stuff out, but I think it leaves some room for someone to try, and their system would probably disagree with Walsh’s a bit.
Walsh mentions turf in his discussion of park factors for arms
Not saying it’s perfect, but he’s aware of it/attempting to control for it.
If a player can get better results through positioning, that should count in his favor.
It’s the same thing with the other fielding factors. A player who does a really good job of positioning himself gets to more balls and makes more outs. That’s a factor in his favor.
It should count in his favor
but probably not as part of his arm rating.
by Jeff Sullivan on Jan 23, 2009 9:59 AM PST up reply actions
That seems like it would often be an unnecessary distinction
If Upton can repeatedly get better-than-average results with his arm by positioning himself well, then his arm has above average value, whether or not he throws hard.
In a similar vein, we can distinguish between control pitchers and power pitchers, but mostly we just want to know who is good at run prevention. Here, we may be able to distinguish between strong throwers and smart throwers, but we mainly just want to know who is going to be best at run prevention.
I agree with you
but if you’re going to create a statistic intended to measure players’ arms, then you should probably limit your focus to their arms.
by Jeff Sullivan on Jan 23, 2009 1:52 PM PST up reply actions
It seems like changing the name to be extremely accurate could make for a cumbersome title
But maybe I’m just not creative enough at the moment.
I've thought about it more and changed my mind
I think you’re right. There’s no need to make a distinction, provided the reader understands that it’s not a raw measure of throwing strength and accuracy.
by Jeff Sullivan on Jan 23, 2009 1:59 PM PST up reply actions
That is
if you were trying to make a statistic intended to measure raw throwing strength and accuracy, then you would want to factor this stuff out if you can, but if you’re just trying to see who saves and allows runs in the field for reasons other than his range, then this is the way to go.
by Jeff Sullivan on Jan 23, 2009 2:00 PM PST up reply actions
If players acted on perfect information, there would be no kills.
If you knew in advance, with 100% certainty, that you were going to be thrown out, why run?
Well the ability to predict the future is a bit more perfect than I'm implying
I’m wondering if the players all knew exactly how good someone’s arm really was (Kill+, or something), then they wouldn’t run more or less than the ‘right’ amount on them. But then again I guess that amount changes a lot based on game situation, etc, sort of like bunting.
There would obviously be a very strong correlation but I don't know that it would be that simple
by Jeff Sullivan on Jan 22, 2009 11:09 PM PST up reply actions
Is Kill+ predictive of the next year's Hold+? Does an increase in Kill+ lead to a lower Hold+?
As players age who had high Kill+ do their Hold+ stay high and their Kill+ decrease? (reputation)
by Edgar for Pres on Jan 23, 2009 5:50 AM PST up reply actions
These are questions that would be better asked of John Walsh
by Jeff Sullivan on Jan 23, 2009 9:58 AM PST up reply actions
Yeah just making sure I didn't miss anything obvious
I know he has put out a couple articles. Wasnt sure if I’ve seen them all.
by Edgar for Pres on Jan 23, 2009 2:25 PM PST up reply actions
Of note
The best way to do this is exactly how the author did this. He can improve the model by incorporating where the ball was hit, what type of hit it was, and how hard it was hit (and the speed of the baserunner).
Lo and behold, that is exactly what I do (other than adjusting for the speed of the base runner which I might add at some point), and my numbers should be up on Fangraphs shortly!
From the same comment
My outfield arm numbers, BTW, are almost exactly the same as these, even with the refinements.
For the record, Walsh’s arm ratings are available on the THT player pages.
by Jeff Sullivan on Jan 23, 2009 12:21 PM PST up reply actions
Oh yeah, what about infield arms? Or catcher arms?
What are the estimates of how many runs Beltre is worth over, say, Eckstein?
This almost seems like overkill
We don’t adjust hit totals for what kind of pitch the pitcher threw, where the pitch was located, how hard the ball was hit, or how the defense was positioned. If we did, it would be like introducing some kind of PrOPS-on-steroids.
MGL’s adjustments may well do a better job than Walsh’s system for measuring how well the outfielder throws, but I’m not terribly convinced that it will do a better job at measuring how productive an outfielder is at preventing runs with his arm.
In theory it should improve the predictive value a little bit
by Jeff Sullivan on Jan 23, 2009 1:54 PM PST up reply actions
I can see the allure
But it seems like you would need to make an extra step or two after the adjustments to make sure that the statistic reflected what you were trying to predict.
Incorporating hit type and location is important unless you figure that every player gets the same distribution of opportunities
that’ll probably be true over a large sample, but if you’re looking at, say, single seasons, then it’s something to consider.
by Jeff Sullivan on Jan 23, 2009 2:03 PM PST up reply actions
Yeah
Now that I’ve thought about it more from the angle of equalizing the distribution of opportunities, those adjustments make more sense.
The one remaining caveat that I can think of is that we know that hold percentage is different for different fielders, so there could be systematic and long-lasting differences between the distribution of opportunities presented to different outfielders. In particular, if you have a high hold percentage, then you may often be throwing against the fastest runners in the league and off the top of my head, I’m not sure how we should deal with that.
Do people give catchers extra credit for catching good basestealers? I’m not as up on my catcher defense reading as I should be. And in general, it seems like the analysis for catchers’ arms should be pretty similar to the analysis for outfielders’ arms.
"systematic and long-lasting differences between the distribution of opportunities presented to different outfielders."
Seems like the best explanation I’ve got for why, say, Kansas City has such a huge impact on these numbers.
Park effects matter, and that’s most of it, but weirdness like this makes me wonder if how systematic the differences might be. Walsh used a TON of data, so it seems like it basically can’t be a distro thing, or anyway a distro thing that’s not entirely caught up in a park effect… I just can’t see how the park would play such a big role.
re: catcher's arms.
It is. And you don’t need to deal with that, because hold percentage is the more valuable (for OF, reverse true for C) of the numbers. That you have a high hold percentage is already dealing with the bias that then the runners who do run on you are probably faster on average.
But if it already deals with that bias correctly...
…then why add MGL’s adjustments?
The underlying logic behind MGL’s adjustments would seem to be that not all fielders have the same kinds of opportunities in a given year and we should try to adjust it so that all fielders are judged in a context where they are given the same kinds of opportunities. But if fielders systematically get different kinds of opportunities, we shouldn’t try to adjust them to that kind of a neutral context.
I don't think there's anything systematic about batted ball type, location, or velocity for defensive outfielders
by Jeff Sullivan on Jan 23, 2009 4:09 PM PST up reply actions
Yeah, probably not
It’s the types of baserunners that I’m mostly worried about.
MGL specifically doesn't adjust for speed of the baserunner,
which is what I am talking about.
This is kind of difficult for me to wrap my head around right now
but wouldn’t you want the adjustment? Why penalize outfielders for having high hold rates? Why reward outfielders to having low ones? Ideally you want everything to be normalized to a common average, because unless I’m thinking of this incorrectly, that’s the only way to compare.
by Jeff Sullivan on Jan 23, 2009 6:35 PM PST up reply actions
I had two problems that I think I have figured out
First off, I was thinking of Kill+ as representing kills per runner attempting to advance rather than kills per play. Which, if I understand correctly now, Kill+ is Kills/(expectedKills) where expectedKills is determined by the number of times a situation occurs rather than the number of times a runner attempts to advance.
Secondly, I was thinking about only adjusting for the speed of runners attempting to reach an extra base. If I think about it from the beginning of the play, then I can see where a runner with speed V on average holds H% of the time, is killed K% of the time, and advances safely the rest of the time—but H% and K% are functions of V. If we do that for each baserunner that an outfielder faces, then we would have an expected H% and K% and we could see how far above/below average that fielder was. As it is now, we just assume that H% and K% is the same for all baserunners. (To within park factors.)
I guess now—and really this is just a minor quibble—it would be interesting to see rankings of the outfielders based on kills/(kills+advances). From a strict value standpoint, kills per situation is what we want to know, but if we want to make statements about whether or not runners are being too aggressive or too cautious against certain outfielders, it seems like kills/(kills+advances) would be good to know.
tangotiger's comment
at the end about only needing 1/2 season of data to draw conclusions about a player’s arm is interesting.
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recent FanPosts
In Association With
Big League Stew - MLB - Yahoo! Sports
- College player makes impossible over the fence catch to save tournament game (Video)
- Security!: Camden Yards guard collides with Royals’ Mike Moustakas
- The Juice: Nelson Cruz matches career high eight RBIs, Pujols stays hot in Seattle
- Benches clear at Fenway as Red Sox-Rays rivalry heats up
- Keeping good company: Albert Pujols becomes fourth-youngest player to hit 450 home runs
