Joe Kelly and the Trap of ERA

Tonight, the Cardinals and Dodgers square off in Game 1 of the NLCS. The Dodgers are throwing Zack Greinke, who you probably know as one of the best pitchers alive. The Cardinals are throwing Joe Kelly, who, if you don’t watch the Cardinals regularly, you may have never heard of. But, being the Cardinals, it is no huge surprise that they have found some moderate prospect in the third round and turned him into an ace. This is what they do. So, Joe Kelly and his 3.08 career ERA is taking the mound for the Cardinals tonight as yet another example of the Cardinals ridiculous player development success.

Except Joe Kelly is a little different than the rest of the terrific young arms the Cardinals keep pulling out of thin air. A table of the most often used 27-and-under pitchers that St. Louis has thrown over the last two years, to illustrate the difference:


Name IP BB% K% HR/9 BABIP LOB% GB% HR/FB ERA FIP xFIP WAR RA9-WAR
Lance Lynn 377 9% 24% 0.71 0.317 74% 43% 9% 3.88 3.38 3.63 6.0 4.4
Joe Kelly 231 8% 16% 0.78 0.297 78% 51% 10% 3.08 4.00 4.12 1.2 3.5
Shelby Miller 187 8% 24% 0.96 0.280 81% 39% 9% 2.94 3.54 3.70 2.5 3.8
Jaime Garcia 177 6% 19% 0.66 0.327 70% 57% 10% 3.81 3.20 3.36 3.2 1.6
Trevor Rosenthal 98 7% 33% 0.55 0.313 77% 47% 8% 2.66 2.18 2.52 2.4 1.6
Michael Wacha 64 7% 25% 0.70 0.275 80% 44% 7% 2.78 2.92 3.36 1.1 1.6

By innings pitched, Kelly ranks second only to Lynn in terms of young pitcher workload for the Cardinals since he was called up last year. And over the course of essentially one full season’s worth of pitching, Kelly’s results have been excellent, as his +3.5 RA9-WAR show. But, note how he got there.

Kelly has the lowest strikeout rate of the group, by a good margin, and has the second highest walk rate. While pretty much every other young Cardinals hurler has dominated the strike zone, Kelly has thrown 231 innings as a pitch-to-contact groundball guy, only he hasn’t really limited walks the way you want a pitch-to-contact groundball guy to do. In terms of generating walks, strikeouts, and ground balls, Kelly has been pretty ordinary.

But that ERA isn’t ordinary, and so now is usually the part of the post where we try to talk about how Kelly has either limited his home runs on fly balls or the amount of hits he allows on balls in play. Except Kelly doesn’t really do either of those things either. His .297 BABIP and 10% HR/FB rates are almost exactly league average, and don’t really explain any of the variance between his excellent ERA and his mediocre FIP and xFIP.

Instead, we come back to sequencing. It’s the 2013 Cardinals, so of course we do. Based on the quantity of individual events Kelly has allowed, he hasn’t really been that great as a big league pitcher. Based on the order in which those events occurred, though, his results have turned out strongly in his favor. Here’s how this has worked over his career.

Split IP BB% K% HR/9 BABIP FIP xFIP AVG OBP SLG wOBA
Bases Empty 119 10% 15% 0.53 0.315 3.81 4.35 0.271 0.344 0.380 0.323
Men on Base 111 6% 16% 1.05 0.275 4.21 3.87 0.251 0.312 0.393 0.305
Men In Scoring 61 8% 17% 0.74 0.224 3.97 4.01 0.200 0.280 0.296 0.250

Basically, Kelly’s .297 BABIP allowed is skewed heavily towards bases empty situations. With runners on base, his BABIP has been a little below the average, and with runners in scoring position, it’s been absurdly low. And if you’re not giving up hits with men in scoring position, you won’t give up many runs either. This has basically been Kelly’s recipe for success in the big leagues: put them on and then leave them there.

Of all the ways to run a low ERA, this is probably the least sustainable. Some pitchers can hold down hits on balls in play, and some pitchers can hold down their home run rate even while giving up a lot of fly balls, but there simply isn’t the same observed population of pitchers who can consistently get into and out of jams on a regular basis. Good pitchers don’t allow baserunners, basically. They don’t all do it the same way, but in general, the pitchers succeed by holding down wOBA, not by redistributing their outs to the situations that matter the most.

Over the last two seasons, Kelly is one of 135 pitchers to throw at least 200 innings. In terms of R/9, he ranks 23rd, right in between Hiroki Kuroda and Cole Hamels. In terms of wOBA allowed, he ranks 62nd, tied with Ricky Nolasco and Bronson Arroyo. The correlation between wOBA allowed and RA9 is 0.9, even in a sample of just a couple hundred innings for each pitcher. Even with all the noise, it is quite clear that wOBA allowed drives runs allowed, and that pitchers can’t really sustain results that different from their expected runs based on wOBA by very much.

While Kelly’s ERA would tell us he’s one of the Cardinals terrific young arms, he’s pitched more like a back-end innings eater than any kind of future ace. He’s not a bad pitcher, but he’s not Michael Wacha, Shelby Miller, Lance Lynn, or Jaime Garcia. He’s Kyle Lohse with more walks, at least in terms of statistical profile. The Cardinals have long had success with guys like Lohse, and he may very well shut down the Dodgers tonight, but it should be pretty clear that Kelly is the team’s worst starting pitcher, even with his particularly shiny ERA thus far.

The gap between Kelly and Shelby Miller isn’t so large that this is clearly the wrong call, especially with the way Miller’s strikeout rate sunk in September, indicating that he might just be out of gas after a long season. But with Miller available out of the bullpen and better starters to come in the next few games, Kelly shouldn’t have a very long leash. The Cardinals shouldn’t get trapped into thinking they’re throwing a 3.08 ERA starter tonight. He’s not that good, and they shouldn’t be afraid to replace him with better pitchers to keep the game close.



Print This Post



Dave is the Managing Editor of FanGraphs.


Sort by:   newest | oldest | most voted
Canard
Guest
Canard
2 years 9 months ago

Maybe Kelly is just ridiculously good at pitching out of the stretch?

Pirates Hurdles
Guest
Pirates Hurdles
2 years 9 months ago

The he would be equally good with a man on first and not just when the runners are in scoring position.

Patrick
Guest
Patrick
2 years 9 months ago

Some dudes pitch up to tense situations. But what is more telling to me is his groundball rate. 51% is darn good. Also, these are pretty small sample sizes.

tehzachatak
Member
tehzachatak
2 years 9 months ago

Some dudes pitch up to tense situations.

Please, provide data to back this up. I am eagerly awaiting your response.

chuckb
Guest
chuckb
2 years 9 months ago

You’re going to be waiting a long, long time.

AK7007
Member
AK7007
2 years 9 months ago

You can’t tell us that good results have come from a bad process and are therefore unsustainable. Blasphemy! The will of the best fans in baseball keep those RISP from reaching home, luck is for losers! Loud Noises Loud Noises!

Ironic
Guest
Ironic
2 years 9 months ago

Haven’t you ever heard of Jack Morris?

BrianB
Guest
BrianB
2 years 9 months ago

The situation called LIFE

witesoxfan
Guest
witesoxfan
2 years 9 months ago

Jack Morris? His career xFIP was 3.94 and his postseason xFIP was 3.74. That’s not a significant difference. And, really, has anybody ever gotten more mileage out of 13 starts?

Antonio Bananas
Guest
Antonio Bananas
2 years 9 months ago

I can’t express how happy I am that cards fans constantly calling themselves “best fans in baseball” becoming a running Internet joke makes me. Yes I’m very bitter.

Billy
Guest
Billy
2 years 9 months ago

You get at something interesting. I’ve always wondered why we don’t use the same stats for hitters and pitchers sometimes. Why is it OBP for hitters, and WHIP for pitchers? WHIP is more or less opposing OBP (since walks and hits are by far the most common ways batters reach base). It makes it easier for us to understand the statistics of both kinds of players compared to each other. Opposing OBP is basically how well you get guys out. Opposing wOBA would also include how much damage they’re doing when they’re not getting out.

We could see a pitcher’s OBP allowed is .250 and say “this guy looks pretty good since he turns the league on average into Brendan Ryan.”

Manifest
Guest
2 years 9 months ago

And why do pitchers have a pitch count for a game, but innings limits for a season?

Jonah Pemstein
Member
Member
2 years 9 months ago

http://www.fangraphs.com/community/wrc-for-pitchers-and-koji-ueharas-dominance/

I basically did what you just said here – found the opposing wOBA for pitchers – and converted it so it’s scaled like ERA.

Shannon
Guest
Shannon
2 years 9 months ago
James
Guest
James
2 years 9 months ago

Knowing how these things usually work out I imagine he’s going to go 7 shutout innings tonight. Baseball.

Mike Green
Guest
Mike Green
2 years 9 months ago

FWIW, Kelly’s minor league career, taken as a whole, is consistent with Dave’s view.

You wouldn’t think that Kelly’s ground ball tendencies would help him that much. Cardinals’ opponents hit .243/.243/.259 when putting the ball on the ground (league averages .235/.235/.254).

chuckb
Guest
chuckb
2 years 9 months ago

Many of Kelly’s innings the last 2 years also have come in relief so it’s not exactly an apples-to-apples comparison to Cole Hamels and Hiroki Kuroda. Kelly’s a decent arm but the only way he should even be starting in this series is if the Cardinals have basically decided to shut Miller down for the season. If then, he should be pitching in game 4, not in games 1 and 5.

AverageMeansAverageOverTime
Guest
AverageMeansAverageOverTime
2 years 9 months ago

Having Wacha, Lynn, Garcia and Miller, the Cards don’t need Kelly to be an ace.

Todd
Guest
Todd
2 years 9 months ago

Except they could be pitching Lynn or Miller now, and they’re not.

semperty
Guest
semperty
2 years 9 months ago

Tonight we do. Miller is the only one of those four available (maybe Lynn, if Miller starts game 4) so we don’t really have room for Kelly to be bad tonight.

chuckb
Guest
chuckb
2 years 9 months ago

There’s a decent chance Miller won’t pitch at all in the series.

Todd
Guest
Todd
2 years 9 months ago

Miller not pitching for some undisclosed reason is all well and good, but what about Lance Lynn? He’s the better pitcher, and even narratively Lynn is supposed to struggle on the road while Kelly does well there, so why not throw Lynn at home?

chuckb
Guest
chuckb
2 years 9 months ago

My guess is because Lynn stunk in game 2 of the Pirates series.

Todd
Guest
Todd
2 years 9 months ago

Yeah, probably… but you’ve gotta pitch him sometime (unless Miller is available), and it’s not like Kelly lit the world on fire in game 3.

chuckb
Guest
chuckb
2 years 9 months ago

Agreed. I was just explaining the rationale, not defending it.

Sean
Guest
Sean
2 years 9 months ago

and if you look at his numbers only as a starter, his ERA gets better and his xFIP gets worse. *shrugs*

BrianB
Guest
BrianB
2 years 9 months ago

My first two thoughts:

1. While Kelly’s peripherals are very average, he throws very hard compared to most starting pitchers (94.9). Obviously most hard throwers get a good amount of strikeouts. For whatever reason he doesn’t.

2. As mentioned above, he’s clearly better out of the stretch. Remember the Salty interview we read a month or so ago? He had a theory that some hitters are much worse when a pitcher is in the stretch. He thought great numbers with RISP and terrible numbers with RISP were not a coincidence. It changes the batter’s timing according to Salty. Maybe something with Kelly’s mechanics in the stretch throws off timing with the batters? Maybe he should pitch out of the stretch all the time?

Mike Green
Guest
Mike Green
2 years 9 months ago

His LD rate is much lower with runners on base. His pop-up rate is higher. This would partially explain the BABIP improvement, and would be consistent with pitching better out of the stretch than the wind-up. He also has a higher GB rate with runners on.

The strangest thing is that he has a much lower walk rate with runners in scoring position than with nobody on. Normally, pitchers will have a much higher walk rate with runners in scoring position, because of the pitch-around with a runner on second and a dangerous hitter at the plate (or to set up the DP with less than two out). Obviously, this doesn’t apply where there is a runner on first as well, but there are many more occasions with runners in scoring position only.

Andrew
Guest
Andrew
2 years 9 months ago

This is an ideal time to cite Occam’s Razor, which is Joe Kelly is worse than his ERA states. Is he better out of the stretch? His K% and BB% are not outside the range of variance. Additionally, these samples are not big enough to make any explanatory statements about his LD% or BABIP rate. As other noted his ERA is lower as a starter, FIP/xFIP higher. He is simply in his current form, as the article stated a back end of the rotation starter.

Jeff Locke sequenced his way to an All-Star selection, and regressed in the 2nd half, Joe Kelly will not sustain a .224 BABIP with RISP.

BigSteve
Guest
BigSteve
2 years 9 months ago

You have to post the “It’s a trap” picture somewhere in the post with that headline.

TanPadreFan
Member
Member
TanPadreFan
2 years 9 months ago
Oh, Beepy
Guest
Oh, Beepy
2 years 9 months ago

Not doing exactly that is what’s keeping this from turning into the Bleacher Report.

Please go back there.

ntr Admiral Ackbar
Guest
ntr Admiral Ackbar
2 years 9 months ago

Actually, it reminds many of us of the days when Baseball Think Factory was still alive…

Matthew Cornwell
Guest
Matthew Cornwell
2 years 9 months ago

I am going on memory here, but I am pretty sure the r=.5 point is about the same number of seasons for LOB% as it is for BABIP. About 7-8 seasons. I think year to year correlation for LOB% is about the same as it is for BABIP and HR/FB too. You can’t tell much about a pitcher’s skill at any of the three events given 1-2 seasons, but I don’t think LOB% is my any less concrete than BABIP or HR/FB.

Matthew Cornwell
Guest
Matthew Cornwell
2 years 9 months ago

MGL over at Tango’s site:

“I run a regression for 05 on 06, 07 on 08, etc.). That is for players with at least 300 PA/TBF in each of both seasons. That suggests a “sequencing skill” that takes something like 4000-5000 PA/TBF before we regress 50%.”

So it looks like 7 full seasons or so. That is about the same as BABIP.
Still, not enough sample size to make any concrete observations regarding Kelly, of course.

Mr Punch
Guest
Mr Punch
2 years 9 months ago

Or perhaps there really is such a thing as “clutch,” but it’s very rare. Just kidding (maybe).

derp
Guest
derp
2 years 9 months ago

If Kelly was such a good pitcher with runners on base, why can’t someone coach him to smarten up and pitch like that all the time or, heaven forbid, just tell him to pith out of the stretch all the time?

This defeats any possible clutch argument that could be made. My ONLY guess as to his success would be how Gerrit Cole pitched so effectively with mediocre whiff rates for most of the year. A 95 MPH fastball is tougher to square up even if it’s thrown right down the middle and the pitcher is not predictable. I could guess that the Cardinals are telling him to attack hitters as if he’s going for a strikeout with no one on, and to just get it in the strike zone once men get on. It’s a good way to teach a pitcher how to command his pitches without opening him up to big innings where he walks everyone and then serves up a meatball. He obviously doesn’t have very good offspeed pitches that are still developing, which is why attacking for strikeouts isn’t working and just causing more walks now.

Matthew Cornwell
Guest
2 years 9 months ago

Just like there are certain approach types that tend to produce better WPA Clutch stats from the offensive side, I am sure there are some approach types that help pitchers too. Not probably widespread, but somewhere in the noise.

The Book showed that pitchers do pitch differently from the stretch. Not necessarily better or worse. Tom Glavine and his preposterous splits is an interesting case study, as we all know.

CircleChange11
Guest
2 years 9 months ago

There are also some athletes that do not focus as well as they do in certain situations that demand it.

You can’t be “locked in” all the time, but you can lock in when you have no choice for it.

We could just as easily say, why does evanser holyfield seemingly fight at his best once he’s been rocked?

It could be adrenaline, a flighty mind, just luck, or just randomness.

The cardinals have had success with the williams-suppan-pineiro-lohse types … but as someone else ponted out, Kelly throws 95. Thats a break from the typical pitch to contact 2-seamer/sinker type.

The Real Slim Schadenfreude
Guest
The Real Slim Schadenfreude
2 years 9 months ago

Some of Kelly’s success is luck-driven, no doubt. But is there any way to see if his velocity is higher out of the stretch, or with RISP? Because, just anecdotally of course, it seems that his 4-seamer can range anywhere from 92-98 from start to start, and even within some starts.

He’s a little guy who cannot throw 95-97 forty or fifty times a game, so maybe he saves a bit of V for the more essential situations — enough to make *some* of the ERA/FIP discrepancy a genuine reflection of talent. I have no idea whatever if there’s any validity to this theory… is there data?

The Real Slim Schadenfreude
Guest
The Real Slim Schadenfreude
2 years 9 months ago

Forgot to add:

Shelby Miller, 1st 25 starts, 140 IP, 151 K’s, 41 BB’s

Next (and final) 6 starts, 34 IP, 18 K’s, 16 BB’s

So, over Miller’s final six outings his strikeout rate dropped by half, and his walks spiked. Yuck. As a St. Louis fan, I’m concerned about structural damage even more than simple fatigue. If it were my decision, I think he’d be done for the year — those splits are just too stark. (As opposed to Koo Stark.)

Eric Feczko
Guest
Eric Feczko
2 years 9 months ago

It is certainly possible, but I think fatigue/dead arm is a more likely explanation.

Shelby has never thrown more than 150 innings in a year. The fact that he dropped off at about 140 innings is a red flag itself.

However, if either explanation were true, we should expect to see some alteration in Shelby’s pitching. The first sign of injury would be a loss in velocity:

http://www.fangraphs.com/pitchfxo.aspx?playerid=10197&position=P&pitch=FA

Here, we see no drop in velocity for Shelby’s fastball. Unusually his changeup and curveball have lost velocity, but I’m not sure whether that really means anything here.

In looking at his game charts:

http://www.fangraphs.com/pitchfxg.aspx?playerid=10197&position=P&season=2013&date=0&dh=0

I don’t see anything as a red flag for the last six games of the season relative to average. His release point, his command, and his movement all seem to be within the norm.

There’s a good chance I may have missed something, but his low strikeout rate may just be a non-predictive fluke. Still, the fact that he was held out for the NLDS suggests that the cards may know something about his health that we don’t. However, if he was seriously injured, the cards would not let him on the roster for the NLCS either.

pft
Guest
pft
2 years 9 months ago

Kelly probably pitches a no hitter tonight.

The thing is in any given game the variance in a pitchers performance can make a league average pitcher look like Cy Young and a great pitcher with awesome career stats look like a dog.

I do agree Kelly is probably not as good as his ERA suggest, but am not sure it means anything for this particular game. The pitchers who handle the pressure of the post season sometimes pitch beyond their true talent level and vice versa.

Rob
Guest
Rob
2 years 9 months ago

First inning: Vintage Joe Kelly. Let two guys on second and third with one out, then start pitching for real and work out of it.

CatheyBarrett52
Guest
CatheyBarrett52
2 years 9 months ago

Start working at home with Google! It’s by-far the best job I’ve had. WORK LESS EARN MORE Exit35com

Farley Facetious
Guest
Farley Facetious
2 years 9 months ago

I love, love , love your offer, Cathey! And this is exactly the right place to advertise your wondrous work-at-home breakthrough! Tell us all more, so much more about how we can defy conventional logic and earn more while working less! Or email me personally at

CatheyBarrettyoulyingsackofexcrement@msn.com

Thanks, Cathey!

Nate
Member
Nate
2 years 9 months ago

How does one view pitchers ranked by opposing wOBA? I can see it under splits (e.g., vL, vR), but not in the overall leader boards.

Thomas Au
Guest
Thomas Au
2 years 9 months ago

Kelly is very “clutch.” He “turns it on” and all his metrics improve with men on base, except his home run rate. This happens because he is “trying harder.”

Thomas Au
Guest
Thomas Au
2 years 9 months ago

Kelly is very “clutch.” He “turns it on” and all his metrics improve with men on base, except his home run rate. This happens because he is “trying harder.”

He’s a former reliever and can’t afford to throw his hardest all the time. But does so when it counts most.

wpDiscuz