Exceptional Defense Touches Everyone
Here’s something that should be pretty evident: If you’ve got a ground-ball pitcher, you want him pitching in front of a strong infield defense. Likewise, if you’ve got a fly-ball pitcher, you want him pitching in front of a strong outfield defense. I feel like I don’t even need to explain the thought processes. How many times did people express concern over Rick Porcello starting for last year’s Detroit Tigers? Porcello’s a ground ball guy. Last year’s Tigers started Miguel Cabrera and Prince Fielder at the corners. Intuitively, that could’ve been a problem.
OK. As presented on FanGraphs, the UZR era stretches back to 2002. Over that span, last year’s Tampa Bay Rays had one of the best infield defenses, at +50 runs. Not surprisingly, ground-baller Alex Cobb posted an ERA well below his FIP. More surprisingly, fly-baller Matt Moore showed an even bigger positive difference. Let’s flip things around. The 2004 New York Yankees had one of the worst outfield defenses, at -68 runs. Not surprisingly, fly-baller Javier Vazquez pitched below his peripherals. More surprisingly, ground-baller Jon Lieber showed an even bigger negative difference. These are just carefully selected individual examples, but they help to set up a bigger-picture study.
Here’s the handwritten note on my notepad, from a few nights ago:
GB pitchers and great/bad infields
It started with me wanting to know how ground-ball pitchers have been affected by pitching in front of great infields and lousy infields. That turned out to be just the first step, as the study grew from there. It only made sense to examine all four of the following:
- Ground-ball pitchers in front of good/average/bad infields
- Fly-ball pitchers in front of good/average/bad infields
- Ground-ball pitchers in front of good/average/bad outfields
- Fly-ball pitchers in front of good/average/bad outfields
We can all guess the results, but we might as well get the actual numbers. So I went after the actual numbers, looking at data covering the past dozen years. I got information for a whole bunch of teams, and I got information for a whole bunch of players.
I generated a spreadsheet that included every pitcher season with at least 100 innings. I deleted those pitchers who were moved midseason, which left me with a sample of 1,578. For each pitcher season, I calculated a z-score for the ground-ball rate. I defined ground-ball pitchers as guys whose ground-ball rates were at least one standard deviation above the mean. I defined fly-ball pitchers as guys whose ground-ball rates were at least one standard deviation below the mean.
The next step was pairing players with corresponding team defenses, split into infield and outfield. I messed around with z-scores a little bit more. I defined a good infield or outfield defense as a unit that was at least one standard deviation above the mean. I defined a bad infield or outfield defense as a unit that was at least one standard deviation below the mean. For infields, one standard deviation was 21.4 runs. For outfields, one standard deviation was a nearly identical 22.5 runs. All the other defenses, by the way, were included in the “average” group. So this is an examination of groundball pitchers and fly-ball pitchers with good infields, average infields, bad infields, good outfields, average outfields and bad outfields.
I know this probably seems like a lot. The results make it out to be a lot simpler. There are going to be four tables, and here’s the first of them. This is data for ground-ball pitchers, with various infield defenses.
| Inf. Defense | WAR/200 | RA9/200 | ERA- | FIP- | BABIP | Infield, z | Outfield, z |
|---|---|---|---|---|---|---|---|
| Best | 2.6 | 3.4 | 90 | 95 | 0.285 | 1.4 | 0.0 |
| Average | 2.7 | 3.0 | 93 | 95 | 0.293 | -0.1 | -0.2 |
| Worst | 2.7 | 2.1 | 100 | 96 | 0.303 | -1.4 | -0.3 |
To walk you through real quick — groundball pitchers in front of the best infield defenses averaged 2.6 WAR per 200 innings. They averaged 3.4 RA9-WAR per 200 innings. They averaged an ERA- five points lower than their FIP-, and they posted a .285 average BABIP. The best infields were an average of 1.4 standard deviations above the mean, in UZR. Those same teams also had average outfields.
Nothing about this table should be surprising. Ground-ball pitchers in front of good infields allowed a BABIP 18 points lower than ground-ball pitchers in front of bad infields. The good-infield group outpaced its WAR/200 by 0.8, while the bad-infield group undershot its WAR/200 by 0.6. The message: ground-ball pitchers benefit from guys who are better at dealing with ground balls. Of course.
But now let’s look at fly-ball pitchers, with various infield defenses.
| Inf. Defense | WAR/200 | RA9/200 | ERA- | FIP- | BABIP | Infield, z | Outfield, z |
|---|---|---|---|---|---|---|---|
| Best | 2.4 | 2.9 | 97 | 103 | 0.276 | 1.4 | 0.4 |
| Average | 2.1 | 2.1 | 105 | 106 | 0.285 | -0.1 | 0.2 |
| Worst | 1.9 | 1.4 | 110 | 108 | 0.289 | -1.6 | 0.0 |
Even though you expect ground-ball pitchers to benefit the most from having a good infield defense, it’s not like fly-ball pitchers are left untouched. The good-infield group here outpaced its WAR/200 by 0.5, while the bad-infield group undershot its WAR/200 by 0.5. There’s a 13-point spread in BABIP. You can’t ignore the last column; there’s a small outfielder effect, here. But it’s mostly about the infield. Fly-ball pitchers also benefit, sometimes rather significantly, from guys who are better at dealing with ground balls.
Now to the third table. This is data for ground-ball pitchers, with various outfield defenses.
| Outf. Defense | WAR/200 | RA9/200 | ERA- | FIP- | BABIP | Infield, z | Outfield, z |
|---|---|---|---|---|---|---|---|
| Best | 2.8 | 3.1 | 92 | 93 | 0.292 | -0.2 | 1.6 |
| Average | 2.6 | 2.8 | 95 | 96 | 0.293 | -0.1 | -0.1 |
| Worst | 3.1 | 2.9 | 94 | 92 | 0.301 | -0.5 | -1.4 |
The good-outfield group here outpaced its WAR/200 by 0.3, while the bad-outfield group undershot its WAR/200 by 0.2. There’s an eight-point BABIP spread. Clearly, groundball pitchers derive a bigger benefit from a good infield than from a good outfield, but a good outfield does still help. A really good outfield can help by maybe half of a win.
Now for the last table: fly-ball pitchers, with various outfield defenses.
| Outf. Defense | WAR/200 | RA9/200 | ERA- | FIP- | BABIP | Infield, z | Outfield, z |
|---|---|---|---|---|---|---|---|
| Best | 2.1 | 2.6 | 99 | 106 | 0.278 | 0.2 | 1.5 |
| Average | 2.2 | 2.1 | 104 | 105 | 0.284 | 0.1 | 0.1 |
| Worst | 2.1 | 1.7 | 108 | 107 | 0.287 | -0.2 | -1.3 |
The good-outfield group here outpaced its WAR/200 by 0.5, while the bad-outfield group undershot its WAR/200 by 0.4. There’s a nine-point BABIP spread. Obviously, fly-ball pitchers have gotten some help from good fly-ball catchers. But here something interesting: At least based on this evidence, fly-ball pitchers have been helped or hurt as much by the infields as by the outfields. Also interesting: At least based on this evidence, ground-ball pitchers have been helped more by good infields, relative to bad, than fly-ball pitchers have been helped by good outfields, relative to bad. It could be that I’m seeing something that isn’t there. It could be this would all go away with bigger sample sizes. For the moment, it’s something to think about.
And that’s maybe getting a little too particular. Here’s the most general point: Don’t forget that ground-ball pitchers allow balls in play in the air. Don’t forget that fly-ball pitchers allow balls in play on the ground. When you have a ground-ball guy or a fly-ball guy, it’s easy to pretend they only allow one thing. But infields don’t make a difference only for ground-ball guys, and outfields don’t make a difference only for fly-ball guys. If you’ve got a great infield or a lousy outfield, everyone’s going to feel it. After all this information, it seems so obvious. It is obvious. But defense matters, for everybody — always. It doesn’t matter the pitcher’s specialty, or the defense’s.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
Defense always matters clearly, because fly ball pictures have ground balls, too.
But where is the BABIP difference coming from? Can we see a GB/FL/LD breakdown? Perhaps even an infield/shallow/outfield fly.
This leads me to speculate that infielder vs outfielder positional adjustments may one day be tweaked to favor infielders defense as being more important.
Given the wide variety of ballpark designs, it seems outfield placement would enable infield placement, particularly in places like AT&T. I wonder if, as a first approximation, there is away to contain park factors into the evaluation?
“The 2004 New York Yankees had one of the worst outfield defenses, at -68 runs. Not surprisingly, fly-baller Javier Vazquez pitched below his peripherals. More surprisingly, ground-baller Jon Lieber showed an even bigger negative difference”
The NYY infield last year also had 600 innings of shortstop at a -40/150 UZR, with their top innings for other IF positions all at about average.
SO I don’t see it all that surprising.
I’m confused as to how your calculating your outfield runs though. Looking at the team page, I get this for the highest inning totals in the OF:
Brett Gardner CF: 1166 Inn -.3 UZR/150
Ichiro Suzuki RF: 993 Inn 17.8 UZR/150
Vernon Wells LF: 623 Inn 8.0 UZR/150
Alphonso Soriano LF: 416 Inn 11.5 UZR/150
How is that an outfield that gave up 70 runs defensively. It looks average in CF and great at the corners. What the hell numbers are you using?
I apparently missed the “2004” part.
If only Javier Vazquez was still pitching for the Yankees! Mind you, he would probably still be in the fight for the 5th spot in their rotation.
Up-vote ^
Or to put it more simply, infields are more important as a whole to team defence than outfields. The defensive spectrum features at the left end shortstops, second basemen, centerfielders and third basemen reflecting the primacy of infield defence.
There’s a couple of things here that concern me:
1. Defining “fly ball pitchers” as guys who have one standard deviation less groundballs than average – you haven’t picked guys who are flyball pitchers, you’ve picked guys who don’t get a lot of groundballs. That sounds the same, but it isn’t. You’re also pulling in “guys who just got rocked and gave up lots of LD”
I also worry about the validity of some of these groups where we’ve got WAR/200 differences. Fangraphs WAR is FIP based, and therefore shouldn’t have much difference based on defense (IF FIP is doing what its supposed to, which it doesn’t) You’d be much better off using SIERRA based WAR, or something along those lines.
What would be interesting would be to take your samples, and see what the LD/FB/GB distribution is infront of these defenses. Are the good infields getting more GB hit to them?
I think the definition also neglects to take into account the strikeout tendencies of pitchers.
That too.
I’d be much more comfortable with something along the lines of “Flyball pitchers are guys who get fly balls in excess of 35% of plate appearances” or something along those lines.
I’d have no problem seeing a low strikeout guy fall into both “Ground ball guys” and “fly ball guys” if say he struck out 10% of batters, gave up 50% GB, and 40% fly balls.
I think that the point of this is to see how much benefit given pitcher types receive from differently skilled defensive units. In that case, comparing two WARs, one defense independent (FIP) and one defense dependent (RA9) makes perfect sense.
My point is that if you’re getting different FIPs, you’re (if FIP works right) getting different QUALITIES of pitcher.
There’s no reason to believe that an elite flyball pitcher should get the same advantage from good (whatever) defense as a mediocre flyball pitcher. IE, the WAR/200 spreads cast doubt that the samples are equivalent.
That being said, I think FIP is terrible, so Its probably meaningless (which unfortunately makes the whole study meaningless)
Gotcha. That is very true. You’d would want the FIPs to be similar across groups. For some of the groups they are and for some they are not.
Wait, you think FIP is terrible but you like SIERA? Could you elaborate on that?
FIP is significantly less predictive than ERA for pitchers that pitch in hitters parks. SIERRA doesn’t have that issue.
FIP thinks that a hit is a better outcome than a non-K out for any pitcher with a FIP below 3.2(or whatever the league constant is that year). SIERRA doesn’t have that problem.
FIP’s denominator is IP, which is OUTS/3, which is inherently defense dependant (as the majority of OUTS are made by the defense). SIERRA’s term denominators are all PA.
I’m not a huge fan of Sierra, but it fixes a lot of the obvious flaws with FIP, which is only marginally better than ERA.
“FIP’s denominator is IP”
What changes when you switch it to PA? If you don’t know the answer, you shouldn’t be complaining. If you do know the answer, you know that basically nothing changes.
Plenty of things to complain about where FIP is concerned. That isn’t one of them.
Arc, the denominator of FIP being IP means its not defense independant. Changing it to some combination of K%, BB%, and HR% instead of K/9, BB/9, HR/9 would remove the issue where it penalizes pitchers with FIPs lower than 3.2 for getting outs from non-Ks. It would also make it drastically more predictive. It would also be fielding indepenant, so the actual name wouldn’t be a misnomer.
FIP is a bad stat.
RC, no, that’s exactly what I just asked you about. Your theory that changing the denominator from IP to PA would make the stat more predictive is patently false. Why haven’t you just tried it? Plenty of others have, including myself and of course Tom Tango, FIP’s creator.
When you change the denominator to PA, the results are virtually identical. So go ahead and do it (it’s easy) if it makes you feel better about the purity of fielding independence, but it won’t change anything in terms of the results of predictiveness.
“When you change the denominator to PA, the results are virtually identical.”
My results, and SIERRA both say otherwise.
K% correlates better year to year than K/9. BB% correlates better year to year than BB/9. K/BB is just a useless mess.
This is really simple stuff
Good point re: low grounders vs. high flies. Shouldn’t have done that.
Your part about “guys who just got rocked and gave up lots of LD” brings up something interesting. How do the various kinds of defenses do against line drives? Even good pitchers allow hard contact sometimes, and it would also tell us how much a good defense helps to protect bad pitchers.
In Soviet Russia, ball finds you.
That’s a good conclusion at least, “GB pitchers” tend to add a mere 5% to their ground balls over average, there’s just a handful who go up to 10% or more. One out of twenty balls in play switching from a FB to a GB isn’t insignificant, but there are so many factors that make a pitcher what he is, it used to get way too much attention.
Using WAR/200 to evaluate pitchers may overlook how much harder it is to get to 200 with a bad infield (or outfield). Pitchers whose defenses let on extra runners are not going to go as deep into games.
This gets at some interesting points.
I’d generally see much more work being done with % based stats with a denominator of PA, rather than IP. Innings just brings in way too many other factors
(Innings as a denominator is basically making Outs your denominator. Which means that with anything like Ks, GBs, FBs, etc, your denominator and numerator are in affect partially the same variable. There are situations in FIP, because the denominator is IP, where a pitcher giving up a hit is better than getting a non-K out, and that’s lunacy.)
Yeah, great, so why haven’t you just done it and told us your results? Or at least googled to find out what the results are when others have done it? Tom Tango has offered this for years now.
Had you done either of these things, you’d know that changing the denominator from IP to PA results in no significant changes in the stat. You get the intellectual purity of PA, which I do think is good, but the results are essentially identical.
I have, and I’ve found exactly the opposite. All % based stats correlate better than /9 stats.
But WAR/200IP is simply a normalisation of WAR to how many innings the pitcher pitched. You never have to get to 200IP.
Normalising that way captures the marginal performance advantage over a replacement level pitching.
http://blog.zap2it.com/frominsidethebox/true-detective-poster-art.jpg
touch defense and defense touches you back
I love this analysis.
I’m surprised no one has pointed this out yet, but the reason why groundball pitchers are helped more by infield defense than fly ball pitchers are by outfield defense seems simple to me. Pitchers in general allow more ground balls than fly balls.
The most extreme ground ball pitchers induce ground balls nearly 60% of the time, and allow flyballs less than 25% of the time. On the other hand, the most extreme fly ball pitchers still allow ground balls around 35% of the time, and flyballs less that 50% of the time. So fly ball pitchers still are not allowing as many flies as ground ball pitchers are grounders, meaning ground ball pitcher should experience a wider swing: more value added by good infield defenses and more value lost from bad ones.
Another reason I think is that more balls in the air are the type of balls that basically are not affected by defense. A large portion of fly balls are so routine that no defender could reasonably be expected to mess them up. Most line drives also are basically guaranteed to be hits, or outs if they are hit right to a fielder. A fielder has a very short time to react, meaning a good fielder can’t cover that much more ground than a good one. On the other hand, a routine ground ball is more likely to go for a hit if it happens to be hit to a hole. There are more balls where the difference between a good fielder and a bad fielder is significant.
I guess this just goes to ask, how much more importance is infield defense and what level should teams weight infield defense.
Considering there are 2 middle defenders in the infield, SS and 2B, and one middle defender in the outfield, CF, do we place a 2-to-1 weight factor on the infield.
In which case, is defense on a per player-basis, is still the same. It’s just that there’s more critical defensive positions in the infield.
Thanks for your continuing efforts to introduce the neutral zone trap into baseball.
It’s hard to believe how good the Royals defense was last year. It’s also equally as difficult to believe how poor the Mariners and Phillies D was.
Hi Jeff, very interesting article. Left me w/ a few questions:
1. what is the minimum and avg GB% for the ‘groundball pitchers’ and likewise for the ‘flyball pitchers’?
2. what is the relative number of GBs vs. FBs in play? Seems like having more grounders to field helps the great infielders have a bigger impact. Along the same lines, what is the defensive WAR accumulated by the best infield vs. the best outfield defenders?
I guess I’m wondering if the GB-heavy pitchers are really all that skewed to GBs, and if the infields have a bigger effect because of more chances, or because of a bigger spread in talent among infielders than among outfielders
Her name was actually Mrs. Johnson, she lived next door, and she told me I was special.
Thanks.
Not Mrs. Robinson?
Well, that certainly would’ve been a better joke. 🙁
Ditto what everyone is saying. GB pitchers tend to be low K guys, so they still allow lots of FB. FB pitchers tend to be high K guys and thus allow few fly balls even though they are fly ball pitchers (assuming your definition of FB and GB pitchers are the ratios or percentages of all batted balls).
You don’t say what you mean by “ground ball rates” in classifying pitchers as GB or FB pitchers, but if you didn’t, you definitely want to use GB per 9 innings and FB per 9 innings. Of course once you determine that one group actually allows more FB per game (9 inn) and the other group more GB per game, it becomes obvious or even tautological what the effects of a good or bad outfield or infield will be.
Also, FB’s allowed by FB pitchers tend to be easy ones, FB’s by GB pitchers are harder, and vice versa for GB’s and GB and FB pitchers.