Cardinals SPs - ERA vs. xERA

[mattmitchl44]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

[sikeston bulldog2]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

What does this say? I see McG is twice his era and more, then May with a positive difference. That seems like bookends and quite the extremes.

[scoutyjones2]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

What does this say? I see McG is twice his era and more, then May with a positive difference. That seems like bookends and quite the extremes.

xERA, or Expected Earned Run Average, is a baseball metric that estimates a pitcher's ERA based on the quality of contact and outcomes they allow, independent of defense and luck.

Very little of any of these metrics actually are predictive of ERA. The best of the bunch, SIERA, only explains 20.4% of the variance in subsequent-season ERA. In fact, you’re better off using K-BB% to predict future ERA than any of these ERA indicators.

[Melville]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

What does this say? I see McG is twice his era and more, then May with a positive difference. That seems like bookends and quite the extremes.

It says, in part, that Gorman, Winn, Wetherholt, and Burleson are one of the best infields in MLB just as I predicted - each elite or well above average defensively.
Scott was of course expected to be good in CF, but Walker's continuing growth, and the massive upgrade of Church over Mootbaar in LF is also a factor.

[rockondlouie]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

What does this say? I see McG is twice his era and more, then May with a positive difference. That seems like bookends and quite the extremes.

xERA, or Expected Earned Run Average, is a baseball metric that estimates a pitcher's ERA based on the quality of contact and outcomes they allow, independent of defense and luck.

Very little of any of these metrics actually are predictive of ERA. The best of the bunch, SIERA, only explains 20.4% of the variance in subsequent-season ERA. In fact, you’re better off using K-BB% to predict future ERA than any of these ERA indicators.

Bingo

[Wattage]

McGreevy - 2.18 ERA vs. 5.08 xERA
Liberatore - 4.07 ERA vs. 5.50 xERA
Pallante - 4.34 ERA vs. 4.47 xERA
May - 4.85 ERA vs. 4.64 xERA
Leahy - 4.93 ERA vs. 5.21 xERA

FWIW - over 2021-2025 of 141 pitchers who threw at least 400 IP (which basically limited the list to almost all SPs), an ERA of ~0.7 less than xERA was about the best you can find in terms of positive deviation.

im guessing the reason liberatores xera is so high is because of all the homeruns. im thinking the homeruns are partially a blip and he will get them in control moving forward.

[Jatalk]

Anytime I see an X in front of a stat I know I’m not smart enough to figure it out.

[hugeCardfan]

Could someone rationally explain to me how McGreevy's xERA is 5.08?

[hugeCardfan]

Since I got no response from my buddies at CT, I asked chatgpt to weigh in:

A baseball question. The Cardinals pitcher McGreevy has an ERA of 2.18 a whip of .86 and a BAA of .184 after 8 starts this year. Still, the xERA, or expected ERA for McGreevy, is over 5. Can you explain this to me?

Excellent baseball question.

This is exactly where modern analytics can seem to insult what your eyes are telling you.

For Michael McGreevy:

ERA = what actually happened
xERA = what Statcast thinks should have happened based on contact quality + strikeouts + walks
So when you see:

2.18 ERA
0.86 WHIP
.184 BAA
but 5+ xERA
…the model is effectively saying:

"This run prevention looks unsustainably fortunate."

Why?

What xERA cares about
xERA heavily weights:

strikeouts
walks
exit velocity allowed
launch angle
hard-hit contact profile
It asks:

“Given the quality of contact allowed, how many runs would pitchers usually allow?”

Not:
“How many actually scored?”

Why McGreevy’s xERA may hate him
McGreevy is not a classic overpowering bat-misser.

If his profile looks like:

modest strikeout rate
lots of balls in play
some hard contact that happened to find gloves
strand-rate luck
good defensive support
then xERA says:

"Nope. Regression coming."

Example:

A 104 mph line drive caught by the CF?

ERA: zero damage
xERA: "that was a rocket; hitters usually get rewarded"
Do that repeatedly and xERA climbs fast.

Why WHIP and BAA can still look elite
Because McGreevy may be succeeding through:

command
weak sequencing
inducing poor swing decisions
changing timing
getting hitters off barrel despite modest velo
Models don’t perfectly capture deception, pitch tunneling, command artistry, or sequencing.

That’s where analytics can underrate certain pitchers.

Classic examples:

Kyle Hendricks in his prime
some versions of Adam Wainwright
soft-contact specialists
The model tends to trust strikeouts more than "pitchability."

Small sample warning
8 starts is nothing.

xERA can swing wildly early because:

one bad contact cluster
weird BABIP
small innings sample
So a 5+ xERA this early does not mean McGreevy is secretly terrible.

It means:

"His underlying peripherals do not currently support a sub-2 ERA."

Very different statement.

The baseball translation
If I were scouting this:

Optimist says:
"McGreevy knows how to pitch, induces weak enough contact, changes speeds, competes, and beats models."

Skeptic says:
"He isn't missing enough bats. The contact profile says trouble ahead."

Truth is often in between.

My suspicion?

A true talent ERA might be more like 3.80–4.40, not 2.18 and not 5.20.

If you want, we can pull apart his actual strikeout %, walk %, hard-hit %, barrel %, BABIP, and strand rate and see whether xERA is being unfair or correctly skeptical.

ok
Let’s dissect Michael McGreevy the way a front office would.

The key question:

Is McGreevy beating the model because he has a sustainable skill set, or is he surviving on smoke and mirrors?

The numbers I’d want to inspect (and how to interpret them):

1. Strikeout rate (K%)
This is the first place xERA gets suspicious.

Modern MLB average starter:

roughly 21–23% K rate
If McGreevy is:

16–18% → warning
below 15% → major warning
Why?

Because low-K pitchers allow lots of balls in play, which creates volatility.

Analytics basically say:

“Eventually baseballs find grass.”

2. Walk rate (BB%)
This can rescue a lower-K pitcher.

Good:

under 7%
Excellent:

under 5%
If he pounds the zone, that helps offset fewer strikeouts.

3. Hard-hit %
Huge one.

Statcast “hard hit” = 95+ mph.

League average:
~38–39%

If McGreevy is:

under 35% → good
40%+ → xERA concern justified
Because hard contact eventually hurts.

4. Barrel %
This matters even more.

A “barrel” is ideal damage contact.

League average:
~7–8%

If he’s allowing:

4–5% → strong
8–10% → danger
5. BABIP (Batting average on balls in play)
This often exposes luck.

League average:
~.290–.300

If McGreevy is sitting:

.220 or lower
then some regression is likely.

Not always—but usually.

6. Strand rate (LOB%)
How many runners never score?

League average:
~72%

If he’s running:

85%+
that’s often unsustainable.

Classic “ERA prettier than reality.”

7. Home run/fly ball rate
If he’s allowing hard contact but no homers?

That can normalize fast.

My guess before seeing the sheet
I suspect McGreevy looks something like:

low strikeouts
excellent control
suspiciously low BABIP
high strand rate
middling hard-hit profile
That would create:

Traditional line:
"Wow, ace!"

Analytic line:
"Regression grenade."

The bigger baseball question
Some pitchers consistently beat estimators.

Examples:

prime Kyle Hendricks
some Wainwright seasons
crafty command/changeup guys
Why?

Because models still struggle with:

pitch sequencing
deception
weak-contact management
disrupting timing
command to edges
So xERA is useful—but not gospel.

If you want the real answer, I can pull his current Statcast profile and tell you whether the model is being fair or stupid.[/quote]

Yes, please.

I pulled the Statcast profile for Michael McGreevy, and now the xERA makes perfect sense.

The short version:

The model thinks he’s getting away with murder.

Here’s why.

The actual underlying profile
The good
Avg Exit Velocity: 87.6 mph
That’s actually solid.

League average is roughly ~89 mph.

So hitters are not consistently crushing him.

Hard-Hit %: 34.8%
Also pretty good.

League average:
~38–39%

Again, positive.

Walk avoidance
You correctly noticed the WHIP.

He doesn’t beat himself much.

That matters a lot.

The ugly (this is why xERA screams)
Barrel %: 11.4%
This is the smoking gun.

League average pitcher:
~7–8%

11.4% is bad.

Really bad.

A barrel is the exact type of contact that becomes:

doubles
bombs
extra-base damage
Statcast sees this and says:

“A pitcher allowing this many premium contacts should be giving up more runs.”

xwOBA vs actual results
Actual wOBA allowed:
.269

Expected xwOBA:
.395

That gap is enormous.

Translation:

Hitters should be doing far more damage than they actually are.

That’s usually:

fortunate sequencing
balls hit right at defenders
warning-track outs
excellent defense
plain luck
BACON (balls in play)
Batting average on contact:
.243

Expected:
.349

That is a gigantic gap.

Baseball translation:

“When hitters make contact, outcomes have been dramatically kinder than expected.”

This is exactly what inflates xERA.

Why the numbers look contradictory
At first glance:

"Wait… if hard-hit % is good, why is barrel % awful?"

Because McGreevy’s contact profile may be bimodal:

Lots of:

soft/manageable contact
Mixed with:

occasional extremely dangerous mistakes
The model hates mistake pitches.

My baseball read
This does NOT mean McGreevy is a fraud.

It means he profiles like a classic command-and-pitchability starter who:

limits traffic,
works efficiently,
but lacks overpowering stuff.
The question is whether his sequencing skill is real.

Examples of guys who outperformed models:

Kyle Hendricks
late-career Wainwright
crafty command pitchers
But…

11.4% barrels is hard to dismiss.

That usually catches up.

My honest estimate
Not a 2.18 ERA true talent.

Also probably not a 5.00+ disaster.

My gut:

3.60–4.25 starter if command holds.

That’s still a perfectly useful rotation piece.

Cardinals translation:

“Solid #4 starter who can occasionally look like a #2 when sequencing runs hot.”

That’s valuable.

To push this further, for Melville's benefit, I asked:

So, I presume this is a fielding independent stat

Exactly—mostly yes, but with an important nuance.

xERA is not purely fielding independent.

That’s the subtle distinction.

Think of the family tree:

Pure fielding independent stats
These try to ignore defense entirely.

Examples:

FIP
Uses only:

strikeouts
walks
hit batters
home runs
Formula basically says:

“Ignore everything once the ball is put in play.”

So yes—very defense independent.

xFIP
Same concept, but normalizes HR rate.

Basically:

“Even home run rate can be noisy.”

Even more estimator-driven.

xERA is different
Statcast xERA says:

“We do care what happened when the ball was hit—but only based on contact quality, not actual result.”

It uses:

strikeouts
walks
exit velocity
launch angle
contact profile
So:

A screaming 108 mph line drive right at the shortstop?

ERA: out, no damage
FIP: doesn't care, ball in play
xERA: "That should often be a hit."

So xERA is:

Defense-neutral-ish
but not fielding independent.

A better term:

contact-quality adjusted estimator

For McGreevy:

Traditional stats say:

“He suppresses hits.”

xERA says:

“No, the results suppress hits. The contact quality suggests hitters should be getting rewarded.”

That’s why it’s such a dramatic disagreement.

Quick cheat sheet:

Stat Uses balls in play? Defense independent?
ERA Yes No
WHIP Yes No
FIP No Yes
xFIP No Yes
xERA Yes (expected quality only) Mostly
My view for a pitcher like McGreevy?

I trust xERA more than ERA in small samples…

…but for command/deception pitchers, xERA can sometimes be too harsh because it struggles to capture:

sequencing skill
pitch tunneling
deception
intentional weak-contact profiles
That’s why baseball still needs scouts and models.

OK, for the coup de gras...

A little pushback here. "Pitcher like McGreevy", what does that mean. McGreevy has 8 different pitches and he controls them very well. Have you any feel for sequencing skills and pitch tunneling, deception or intentional weak-contact profiles?

Fair pushback—and a good correction.

I was making an archetype inference from the statistical profile, not claiming detailed pitch-design scouting certainty. That distinction matters.

If we’re talking specifically about Michael McGreevy, then yes: “pitcher like McGreevy” may be underselling what he actually is.

Let me sharpen this.

McGreevy is not simply a generic soft-toss command guy in the old Kyle Hendricks mold.

What we do know:

deep repertoire (as you note, effectively 8 looks depending on classification)
advanced command
starter’s ability to change shapes/speeds
clear feel for attacking hitters rather than merely surviving

That profile absolutely raises the legitimate question:

Can he intentionally outperform contact estimators?

And the answer is: possibly, yes.

Because xERA has blind spots.

Where your pushback has merit

xERA assumes:

"Quality of contact should normalize over time."

But pitchers with advanced pitchcraft can distort contact quality in ways models only partially capture.

Examples:

1. Sequencing

A pitcher who knows when to:

elevate fastball
bury breaking stuff
backdoor a cutter
front-door sinker
change speed after tunneling similar release points

can create worse swings than raw pitch quality suggests.

Same stuff, different sequence = radically different outcome.

Statcast does not model game-calling intelligence well.

2. Pitch tunneling

If multiple pitches share:

release point
early trajectory
visual deception

the hitter commits before identifying shape.

That produces:

weak rollover contact
jam shots
late swings
poor barrel alignment

xERA sees exit velocity + launch angle after the fact.

It does not directly credit why the hitter was compromised.

3. Contact management as a real skill

This is still debated analytically.

The old orthodoxy:

"Pitchers don’t control contact quality much."

The modern view:
Some absolutely do.

Examples:

Framber Valdez
Logan Webb
prime Waino in certain years

Pitchers with movement/profile deception can systematically induce bad contact.

If McGreevy is doing that intentionally, xERA can overstate regression risk.

But here’s the counterweight

The barrel rate is still uncomfortable.

That’s the one stat I can’t hand-wave away.

If he truly has elite sequencing/tunneling, you'd usually expect fewer barrels—not 11%+.

That suggests one of two things:

A) the sample is noisy (very possible after 8 starts)

or

B) when hitters do identify him, mistakes get punished hard.

My honest revised framing

Instead of:

“McGreevy is a pitchability guy xERA dislikes.”

More accurate:

“McGreevy may possess real command/craft skills that xERA under-credits, but the current contact damage profile still suggests caution.”

That’s a much fairer statement.

Your scouting instinct here is legitimate.

A pitcher with eight offerings and advanced command deserves more nuance than a simple estimator shrug.

So there you have it. The erudite chatgpt, when push came to shove, allowed that xERA is fallible. I wouldn't be looking for a 5+ ERA any time soon.

[NYCardsFan]

Some friendly advice: the fact that many posters are no longer willing to waste their time engaging with you on topics related to advanced analytics (or even basic statistics) is not a sign you’ve “won” the argument; it’s a sign they recognize you’re operating in bad faith (at least on these topics) and aren’t the least bit interested in considering and seriously grappling with their answer(s), only in confirming your priors.

[cosmo.kramer]

I'm sure the people who develop xERA are trying their best, but xERA is garbage

[C-Unit]

Some friendly advice: the fact that many posters are no longer willing to waste their time engaging with you on topics related to advanced analytics (or even basic statistics) is not a sign you’ve “won” the argument; it’s a sign they recognize you’re operating in bad faith (at least on these topics) and aren’t the least bit interested in considering and seriously grappling with their answer(s), only in confirming your priors.

They are (female canine animals, especially dogs)

[An Old Friend]

Some friendly advice: the fact that many posters are no longer willing to waste their time engaging with you on topics related to advanced analytics (or even basic statistics) is not a sign you’ve “won” the argument; it’s a sign they recognize you’re operating in bad faith (at least on these topics) and aren’t the least bit interested in considering and seriously grappling with their answer(s), only in confirming your priors.

I think it’s interesting that even though ChatGPT, in thousands of words, said exactly that which some of us have said far more succinctly, his main takeaway still ended up being “xERA” is fallible.

[mattmitchl44]

A true talent ERA might be more like 3.80–4.40, not 2.18 and not 5.20.

That's what some of us have said.

[AnExParrot]

Some friendly advice: the fact that many posters are no longer willing to waste their time engaging with you on topics related to advanced analytics (or even basic statistics) is not a sign you’ve “won” the argument; it’s a sign they recognize you’re operating in bad faith (at least on these topics) and aren’t the least bit interested in considering and seriously grappling with their answer(s), only in confirming your priors.

All you need know about ol' Mel's ability to judge pitching talent is/was his take on Dakota Hudson. That was some funny, funny stuff.