Stats for The Number — Regression Signals
BABIP, FIP-ERA gap, xwOBA, LOB% — the regression signals that tell you where the market is wrong
You should read this if:
You bet MLB props and want to understand the mental models that drive outcomes.
The Core Insight
"Regression to the mean is not just a concept in baseball — it is a quantifiable, timetabled phenomenon. You know roughly when each stat will stabilize and what it will regress toward."
The MLB Mental Model
BABIP
Batting average on balls in play — .300 league average
Predicts: Pitcher BABIP regresses almost entirely to .300. Hitter BABIP regresses to career norm. THE #1 regression signal.
FIP-ERA Gap
Difference between true talent and results
Predicts: ERA 4.50 / FIP 3.00 = pitcher will improve. ERA 2.50 / FIP 4.00 = pitcher will regress.
xwOBA vs wOBA
Statcast expected vs actual — physics-based
Predicts: Positive gap = unlucky contact, will improve. Negative gap = lucky, will regress.
LOB% vs 72%
Left on base percentage — regresses to 72%
Predicts: 85% LOB = lucky sequencing, earned runs will rise. 60% = unlucky, will improve.
The Regression Signals That Matter
Regression is not vibes in baseball — it's math with a timetable. Each stat below regresses to a known mean at a known pace. Your edge is identifying which hot or cold streak is about to correct and betting the correction before the market re-prices.
BABIP (The Regression King)
Regression signalBatting average on balls in play. For pitchers, it regresses almost entirely to .300 — BABIP against is mostly luck, not skill. For hitters, it regresses toward a career norm that reflects line-drive rate and sprint speed.
Why it matters
BIG-IDEAS.md #2: "The single most important regression signal in baseball." Hitter BABIP needs ~820 PA to stabilize, so most of what looks like a hot/cold streak in April–June is pure BIP luck.
How to use it
Hitter running .380 BABIP with .310 career norm → fade hits overs next week. Pitcher with .250 BABIP-against → buy ER overs and hits-allowed overs. Always regress pitcher BABIP to .300.
| Tier | Value | What it means |
|---|---|---|
| Elite | ±10 pts from norm | Neutral — no signal |
| Great | ±20 – 30 pts | Mild regression signal |
| Average | ±30 – 50 pts | Meaningful regression — act on it |
| Poor | ±50+ pts | Extreme — high-conviction correction bet |
"Pitcher BABIP regresses almost entirely to .300." — BIG-IDEAS.md
FIP-ERA Gap
Pitcher regressionThe difference between a pitcher's FIP (true talent, controlling K/BB/HR) and ERA (results). The book's canonical example is Ricky Romero's 2011 season: 2.92 ERA with 4.20 FIP → 5.77 ERA the next year.
Why it matters
Per BIG-IDEAS.md: the single most reliably predictive signal in sports betting. When the market prices the shiny ERA, you price the FIP-backed reality coming for it.
How to use it
ERA − FIP > 0.5 (ERA worse than FIP) → buy ER unders and K overs; pitcher is better than the line. ERA − FIP < −0.5 (ERA better than FIP) → fade ER unders and K props; regression up is pending.
| Tier | Value | What it means |
|---|---|---|
| Elite | |gap| ≤ 0.2 | ERA and FIP aligned — no regression signal |
| Great | |gap| 0.3 – 0.5 | Modest regression lean |
| Average | |gap| 0.5 – 1.0 | Strong regression signal — trade on it |
| Poor | |gap| > 1.0 | Extreme — canonical market misprice |
"Romero's 2.92 ERA / 4.20 FIP in 2011 → 5.77 ERA the next year. FIP saw it coming." — section_3.md
xwOBA − wOBA Gap
Hitter regressionStatcast's expected wOBA (what the batted-ball quality deserved) minus the actual wOBA. Positive gap = hitter unlucky; negative gap = hitter overperforming.
Why it matters
BIG-IDEAS.md: "The most actionable regression signal" on offense. Bypasses BABIP noise by using exit velo and launch angle directly.
How to use it
Gap > +.030 → buy hits/TB overs on the hitter; regression up is coming. Gap < −.030 → fade hits/TB overs on the hot streak. Applied in reverse for pitchers: low wOBA-against + high xwOBA-against = ER over pending.
| Tier | Value | What it means |
|---|---|---|
| Elite | |gap| ≤ .010 | Results match contact quality — no signal |
| Great | |gap| .020 | Small regression lean |
| Average | |gap| .030 – .050 | Actionable — bet the correction |
| Poor | |gap| > .050 | Extreme mispricing — high conviction |
xBA − BA Gap
Hitter regressionStatcast expected batting average minus actual batting average. Specifically useful early in the season before BABIP stabilizes.
Why it matters
BA doesn't stabilize until ~910 PA (late July). In April/May, xBA is the fastest true-talent hit-rate estimator you have.
How to use it
xBA > BA by ≥ .030 → buy hits overs on next game. xBA < BA by ≥ .030 → fade the hot BA, it's fielder luck.
| Tier | Value | What it means |
|---|---|---|
| Elite | |gap| ≤ .015 | No signal |
| Great | |gap| .020 | Mild lean |
| Average | |gap| .030 – .050 | Actionable regression signal |
| Poor | |gap| > .050 | Extreme — classic buy/fade window |
LOB% vs 72%
Pitcher regressionShare of baserunners a pitcher strands. Regresses hard to the league-average ~72%. The book calls it "the ERA of run sequencing."
Why it matters
LOB% is sequencing luck — do baserunners score at league rate or not? Over a full season it regresses, so extreme values flag imminent ER moves.
How to use it
LOB% > 80% → fade ER unders / buy ER overs on the next start. LOB% < 65% → buy ER unders. Signal is cleanest for starters with 100+ IP.
| Tier | Value | What it means |
|---|---|---|
| Elite | 72% ± 3 | Neutral — no regression signal |
| Great | 72% ± 5 | Mild sequencing tilt |
| Average | 80% or 65% | Meaningful — ER lean pending |
| Poor | ≥ 85% or ≤ 60% | Extreme — high-conviction correction |
HR/FB Regression
Pitcher regressionHome runs divided by flyballs allowed. League average is ~9.5%. The book's BIG-IDEAS.md calls it "the most volatile important stat in baseball."
Why it matters
Year-to-year HR/FB is mostly luck and park. Extreme values (below 6% or above 12%) are screaming regression signals — a pitcher at 5% HR/FB is about to give up more dingers.
How to use it
Switch to xFIP whenever HR/FB is outside 7–11%. xFIP assumes league-average HR/FB and gives you the pitcher's true-talent ER projection.
| Tier | Value | What it means |
|---|---|---|
| Elite | 9.5% ± 1 | Neutral |
| Great | 9.5% ± 2 | Mild tilt — check park context |
| Average | < 7% or > 12% | Regression signal — use xFIP |
| Poor | < 5% or > 15% | Extreme — results about to flip |
Pitcher BABIP (Always Regress to .300)
Pitcher regressionBABIP allowed by a pitcher. Unlike hitter BABIP, pitcher BABIP has essentially no skill component — it regresses almost entirely to .300.
Why it matters
A pitcher with a low in-season BABIP-against is getting lucky on BIP, which has inflated his ERA/WHIP artificially. That luck always runs out.
How to use it
Pitcher at .250 BABIP-against → buy ER overs and hits-allowed overs next 1–2 starts. Pitcher at .330 BABIP-against with otherwise fine FIP → buy ER unders.
| Tier | Value | What it means |
|---|---|---|
| Elite | .300 ± .010 | Neutral — no signal |
| Great | .300 ± .020 | Mild regression lean |
| Average | .260 or .340 | Actionable — bet correction |
| Poor | ≤ .240 or ≥ .360 | Extreme — imminent BIP reversion |
"BABIP takes 820 PAs to stabilize for hitters and never stabilizes for pitchers in a single season." — BIG-IDEAS.md
Hitter BABIP vs Career Anchor
Hitter regressionFor hitters, BABIP has a real skill component (LD%, sprint speed) — so they don't regress to .300, they regress to their *own career norm*. Speedy line-drive hitters can sustain .340+; pull-happy flyball hitters sit at .270.
Why it matters
Treating all hitter BABIP as a single .300 target is the novice mistake. The book's point is to regress to the *right* number, and the right number is career.
How to use it
Deviation from career (not .300) is the signal. .310 for a career-.280 hitter = mild overperformance. .330 for a career-.340 speedster is actually underperformance.
| Tier | Value | What it means |
|---|---|---|
| Elite | Current = career ± 10 pts | No signal |
| Great | Current = career ± 20 pts | Mild lean |
| Average | ± 30 – 50 pts from career | Actionable — bet regression |
| Poor | ± 50+ pts from career | High conviction — luck unwind coming |
Framework in Action: The xwOBA Buy-Low Signal
A hitter has .295 wOBA but .359 xwOBA. Statcast says his contact quality deserves a .359, but he is posting .295 — the gap is fielding luck (hard-hit balls at fielders, line drives caught). This gap closes. It always closes. This is the baseball equivalent of "buy low" — the market prices the .295 reality, you price the .359 true talent.
When to Apply This Framework
- ✓BABIP .050+ above or below career norm = strong regression signal
- ✓FIP-ERA gap > 0.5 = the market is pricing ERA instead of true talent
- ✓xwOBA-wOBA gap > .030 = Statcast says results do not match contact quality
- ✓LOB% above 80% or below 65% = sequencing luck is about to correct
When to Pass
- ⚠️Regression signals conflict with each other — some say over, some say under
- ⚠️Market has already moved to reflect the regression
- ⚠️Insufficient sample for the stat to be meaningful (first week of season)
Key Takeaways
- ✓BABIP is the single most important regression indicator in baseball
- ✓HR/FB rate is the most volatile important stat — use xFIP to normalize
- ✓Regression is not a secondary adjustment in baseball — it is the core of your edge
How DMP Helps
DMP surfaces BABIP deviation, FIP-ERA gap, xwOBA gap, and LOB% signals in the research panel with directional arrows showing regression direction.