Stage 2MLB Framework8 min read

Stats for The Arm — Pitcher Evaluation

The stats DMP uses to evaluate pitchers, and why traditional stats like ERA and pitcher wins mislead

You should read this if:

You bet MLB props and want to understand the mental models that drive outcomes.

Prerequisites: The Arm

The Core Insight

"FIP isolates what pitchers control (strikeouts, walks, home runs). ERA is contaminated by defense, luck, and scorer judgment. The difference between them is one of the most exploitable signals in sports betting."

The MLB Mental Model

1

FIP (use) vs ERA (misleading)

FIP = (13*HR + 3*BB - 2*K) / IP + constant

Predicts: True pitching talent. Excellent: 3.20, Great: 3.50, Average: 4.20, Awful: 5.00

2

K% (use) vs K/9 (less useful)

K per batter faced, not per inning. Stabilizes in ~70 BF.

Predicts: Strikeout probability. Excellent: 27%, Great: 24%, Average: 20%, Awful: 13%

3

xFIP (normalizes HR luck)

Replaces actual HR/FB with league-average rate

Predicts: Whether a pitcher's HR rate is sustainable or will regress

4

Never use: W-L or Saves

Pitcher wins measure run support. Saves measure role, not quality.

Predicts: Nothing about pitching skill — these are team and role stats

The Pitching Stats That Matter

Every pitcher stat below is grounded in the book's framework: strip out defense and luck, isolate true talent, and look for the gap between process (stuff, K%, command) and results (ERA, wins). Gaps are where props get mispriced.

FIP (Fielding Independent Pitching)

ERA estimator

A run-average-scale number built only from the things a pitcher fully controls: strikeouts, walks, hit-by-pitches, and home runs. Scaled to look like ERA so you can compare them directly.

Why it matters

ERA is contaminated by defense, shift positioning, outfielder routes, official-scorer judgment, and pure BABIP luck. FIP removes all of it. The book calls FIP "to ERA what true shooting is to field goal percentage."

How to use it

Use FIP — never ERA alone — as the primary pitcher-skill input for ER and K projections. When FIP is much lower than ERA, that ER prop is mispriced on the over side; when FIP is much higher than ERA, the pitcher is living on defense and is about to regress.

TierValueWhat it means
Elite≤ 3.20Ace-level true talent — ER unders live
Great3.50Mid-rotation plus — solid pitcher-first matchups
Average4.20League-average SP — defer to matchup/park
Poor≥ 5.00Back-end / replacement — ER overs, fade K overs

"FIP is to ERA what true shooting is to field goal percentage." — appendix-charts.md / BIG-IDEAS.md

xFIP (Expected FIP)

Regression-adjusted FIP

FIP, but we replace the pitcher's actual HR/FB rate with the league-average ~9.5%. Because HR/FB rate is the most volatile important pitching stat, this tells you what FIP "should" look like once HR luck normalizes.

Why it matters

A pitcher riding a 5% HR/FB rate is about to give up more dingers; one stuck at 15% is about to give up fewer. xFIP catches both cases before the market does.

How to use it

Whenever HR/FB is below 7% or above 11%, lean on xFIP over FIP for ER projections. Flyball pitchers benefit most from this regression adjustment.

TierValueWhat it means
Elite≤ 3.20HR-normalized true talent is elite
Great3.50Strong underlying skill
Average4.20League-average once HR/FB regresses
Poor≥ 5.00Bad even with league-average HR luck

SIERA (Skill-Interactive ERA)

ERA estimator

A FanGraphs ERA estimator that adds batted-ball profile (GB% vs FB%) on top of FIP's K/BB/HR framework. It rewards pitchers whose batted-ball suppression is a skill, not luck.

Why it matters

For extreme groundball pitchers, FIP understates their ER-suppression ability. SIERA captures the fact that grounders carry lower BABIP damage and feed double plays.

How to use it

Use SIERA (over FIP) when projecting ER props for high-GB arms. A SIERA well below FIP is the sinker-baller's true-talent number.

TierValueWhat it means
Elite≤ 3.20Ace including batted-ball skill
Great3.50Strong SP — GB skill counts
Average4.20Mid-rotation
Poor≥ 5.00Poor even accounting for GB tilt

K% (Strikeout Rate)

Rate stat

Strikeouts divided by total batters faced. This is the direct driver of a pitcher's strikeout prop — project expected BF × K%, adjusted for the opposing lineup's K vulnerability and the park's K factor.

Why it matters

K% stabilizes in ~60–70 batters faced — the fastest of any pitcher skill. That means K props are reliable from early April when nothing else is trustworthy yet. The book calls strikeout props "the cleanest bet in baseball."

How to use it

Every pitcher K prop starts here. High K% vs high-K opposing lineup + large-zone ump + dry-air park = stack the K over. CSW% leads K% — if CSW% is up but K% hasn't caught up, buy K overs before the market adjusts.

TierValueWhat it means
Elite27%+K artist — target K overs aggressively
Great24%Above-average whiff stuff
Average20%Neutral — lean on matchup
Poor≤ 15%Contact pitcher — fade K overs

BB% (Walk Rate)

Rate stat / Command

Walks divided by total batters faced. The book's second-most-stable pitcher skill (~170 BF / 7 starts to stabilize).

Why it matters

Walks inflate pitch counts, which caps IP, which caps K ceilings. High BB% also drives WHIP and ER directly via baserunners. A pitcher with great Stuff+ but a Location+ under 100 walks too many to ever be an ace.

How to use it

High BB% pitcher → IP under, walks-allowed over, potentially ER over. Combine with Zone% — a low-Zone% nibbler is the archetype.

TierValueWhat it means
Elite≤ 4.5%Strike-thrower — go deep in games
Great5.5%Strong command
Average7.7%League-average
Poor≥ 9%Walks-over target; IP-under target

CSW% (Called Strikes + Whiffs)

Leading indicator

The share of a pitcher's total pitches that generate either a called strike or a swinging strike. Think of it as dominance per pitch.

Why it matters

CSW% is more responsive than K% — it stabilizes essentially within a single start. A pitcher whose CSW% jumped 3 points in his last outing is signaling a K% rise before the K counts catch up.

How to use it

Check CSW% trend vs. the K% trend. Rising CSW% + flat K% = K over buy before the line moves. Falling CSW% from a name-brand ace = K under fade before the market adjusts.

TierValueWhat it means
Elite32%+Nasty stuff — K overs live
Great30%Above-average dominance
Average28%League-average
Poor< 25%Hittable — fade K overs

SwStr% (Swinging Strike Rate)

Leading indicator

Swinging strikes divided by total pitches. The purest bat-missing signal.

Why it matters

SwStr% is the leading indicator for K%. It stabilizes faster and flags stuff changes (new pitch, velo bump) before the K counts arrive.

How to use it

Use in tandem with CSW% — if both are up, K over conviction rises. Specifically useful when pitch-mix changes mid-season.

TierValueWhat it means
Elite14%+Top-tier whiff stuff
Great12%Above-average swing-and-miss
Average10 – 11%League-average
Poor< 9%Contact pitcher — K props unsafe

Whiff%

Per-pitch quality

Swings-and-misses divided by total swings. Unlike SwStr%, which uses all pitches as the denominator, Whiff% isolates "when they swung, how often did they miss."

Why it matters

Lets you evaluate specific pitches in the mix. A slider with 40% whiff vs. a lineup that chases sliders is a K prop green light.

How to use it

Pair with pitch mix and opposing lineup's splits by pitch type. Elite whiff on the out-pitch + chase-prone lineup = K over stack.

TierValueWhat it means
Elite30%+Dominant swing-and-miss arsenal
Great27%Plus stuff
Average24%League-average
Poor< 20%Contact-friendly arsenal

Stuff+

Statcast-style model

FanGraphs/Eno Sarris model that grades the raw quality of a pitcher's pitches (velocity, movement, spin axis) versus the league, scaled to 100 = average.

Why it matters

Isolates "how good is the pitch" from "where did they throw it" and "what was the result." A pitcher with 120 Stuff+ and a bad ERA is a future-breakout candidate the market hasn't priced.

How to use it

Gap signal: high Stuff+ + high ERA → K over, ER under reversion incoming. Low Stuff+ + low ERA → fade, his results aren't supported.

TierValueWhat it means
Elite115+Top-10% pitch quality
Great108Plus arsenal
Average100League-average stuff
Poor≤ 90Below-average — command/deception required

Location+

Statcast-style model

Models the quality of where a pitcher locates, independent of what pitch he throws. 100 = league-average command.

Why it matters

Separates command from stuff. Elite Stuff+ with sub-95 Location+ is the recipe for walks and missed-spot homers.

How to use it

Low Location+ → walks-allowed over and ER over even with elite Stuff+. High Location+ on a finesse arm is how they survive despite weak velo.

TierValueWhat it means
Elite105+Pinpoint command
Great102Above-average location
Average100League-average
Poor< 97Misses spots — walks/HR prone

HR/FB (Home Run per Fly Ball)

Regression signal

Of the flyballs a pitcher allows, what share become home runs? League average is ~9.5%.

Why it matters

The book calls this "the most volatile important stat in baseball." It bounces year-to-year on park, weather, and batted-ball luck — which means extreme values are the market's most reliable mispricing.

How to use it

HR/FB above 12% or below 7% → use xFIP, not FIP, for ER projections. Flyball pitchers at 14%+ are due for ER-unders; at 5% they're due for ER-overs.

TierValueWhat it means
Elite≤ 5%Extremely lucky — regression coming up
Great7%Below-avg HR suppression skill or luck
Average9.5%League-average
Poor≥ 13%Unlucky or pitches up — regression down coming

LOB% (Left On Base)

Regression signal

Share of baserunners the pitcher strands. The book frames it as "the ERA of run sequencing" — over a full season it regresses hard to ~72%.

Why it matters

Sequencing luck inflates or deflates ERA without telling you anything about true talent. A pitcher stranding 85% of runners has an ERA about to rise; 60% means an ERA about to fall.

How to use it

LOB% > 80% → ER over lean in upcoming starts. LOB% < 65% → ER under lean. The regression is pure math.

TierValueWhat it means
Elite80%Lucky / high-leverage success — fade
Great75%Above-average sequencing
Average72%League-average — no signal
Poor≤ 65%Unlucky sequencing — buy ER unders

GB% / FB% (Batted-Ball Split)

Profile stat

What share of balls-in-play the pitcher turns into ground balls vs flyballs. League averages are ~44% GB and ~35% FB.

Why it matters

GB pitchers suppress HR and allow a few more singles (.240 BABIP on grounders); FB pitchers are boom-bust with more Ks and more HRs. Park amplifies the effect.

How to use it

FB pitcher at Yankee Stadium, Great American, Coors = stack HRs-allowed overs. GB pitcher at any park = fade HRs-allowed overs regardless of opposing lineup power.

TierValueWhat it means
EliteGB 50%+ or FB 30%-Extreme worm-killer — HR-suppressor
GreatGB 47% or FB 33%Pronounced GB tilt
AverageGB 44% / FB 35%League-average profile
PoorFB 40%+Extreme flyball — HR-prone in wrong park

Framework in Action: FIP vs ERA: Rick Porcello

Porcello posted 4.92 ERA in 2014, then won the Cy Young with 3.15 ERA in 2016. Same pitcher. The difference? His team defensive efficiency went from .672 to .707. FIP would have told you he was the same pitcher both years. ERA fooled everyone.

When to Apply This Framework

  • Every pitcher evaluation — always start with FIP, not ERA
  • FIP-ERA gap > 0.5 = strong regression signal for ER props
  • K% is reliable from early April — use it for K props immediately

When to Pass

  • ⚠️Never use ERA alone to evaluate a pitcher for props
  • ⚠️Never use pitcher W-L record for any analysis
  • ⚠️Never trust early-season ERA without checking FIP

Key Takeaways

  • FIP is to ERA what true shooting is to field goal percentage — it strips out what the player does not control
  • K% and BB% are the most stable year-to-year pitcher skills
  • The FIP-ERA gap is one of the most reliably predictive signals in all of sports

How DMP Helps

DMP shows FIP alongside ERA everywhere, highlights the gap, and uses FIP as the primary pitcher skill input.

Continue Learning