Starting Pitcher Projection Variables: ERA Estimators and DIPS Theory

ERA tells you what happened. DIPS theory asks whether it was going to happen anyway. That distinction — between observed outcomes and pitcher-controlled inputs — sits at the heart of how modern projection systems evaluate starting pitchers, and it explains why two pitchers with identical ERAs can have wildly different forward-looking valuations.

• Definition and Scope
• Core Mechanics or Structure
• Causal Relationships or Drivers
• Classification Boundaries
• Tradeoffs and Tensions
• Common Misconceptions
• Checklist or Steps
• Reference Table or Matrix

Definition and Scope

Defense-Independent Pitching Statistics (DIPS) theory was introduced by Voros McCracken in a 2001 Baseball Prospectus article that landed like a small grenade in the sabermetrics community. The core finding: pitchers have very limited control over what happens to balls once they're put in play by opposing batters. Strikeouts, walks, and home runs — the so-called "three true outcomes" — are essentially transactions between the pitcher and the batter, with fielders irrelevant. Everything else that happens on a batted ball involves luck, defense, and park factors in proportions the pitcher can't reliably repeat.

For fantasy projection purposes, this matters enormously. ERA is the most visible pitching stat in standard scoring formats, but it's also one of the noisiest. A pitcher whose 4.50 ERA traces back to a .340 BABIP (batting average on balls in play) against a league average of roughly .300 is probably not as bad as the ERA suggests — his defense may have been actively terrible, or variance bit him in a bad sequence of weeks. ERA estimators built on DIPS principles attempt to reconstruct what the ERA should have been given the pitcher's actual controllable inputs.

The main ERA estimators used in serious projection work are FIP (Fielding Independent Pitching), xFIP (Expected FIP), SIERA (Skill-Interactive ERA), and ERA+ (park- and league-adjusted ERA). Each makes different assumptions about which variables the pitcher controls and to what degree. The starting pitcher projection methodology section covers how these feed into full-season fantasy valuations.

Core Mechanics or Structure

FIP is the most widely used DIPS-based estimator. The formula, as documented by FanGraphs, is:

FIP = ((13 × HR) + (3 × (BB + HBP)) − (2 × K)) / IP + FIP constant

The FIP constant is recalculated each season to scale FIP onto the same run environment as ERA (typically around 3.10–3.20 in recent MLB seasons). FIP ignores all contact outcomes except home runs, treating HR, BB, HBP, and K as the complete set of pitcher-controlled events.

xFIP replaces actual home runs with an expected home run total derived from the pitcher's fly ball rate multiplied by the league-average HR/FB rate (historically around 10–11%). The logic: home run rates on fly balls are noisy year-to-year for individual pitchers, while fly ball rate itself is more repeatable. xFIP therefore normalizes out one additional layer of variance.

SIERA (developed by Eric Seidman and Matt Swartz for Baseball Prospectus) goes further. It accounts for the fact that strikeout rate and walk rate interact non-linearly — a pitcher who strikes out 30% of batters benefits from each additional K more than one who strikes out 18%, because the high-K pitcher faces fewer balls in play to begin with. SIERA also incorporates ground ball rate, recognizing that ground balls result in fewer runs than fly balls even after controlling for BABIP.

ERA− and ERA+ are park- and league-adjusted versions of raw ERA. ERA+ of 100 is exactly league average; 130 ERA+ means the pitcher allowed 30% fewer runs than a league-average pitcher in the same environments. These are descriptive rather than predictive, but they contextualize raw ERA against changing run environments across seasons.

Causal Relationships or Drivers

The projection inputs that most directly move ERA estimators fall into three tiers of pitcher control, as supported by research across multiple seasons of MLB data published by FanGraphs and Baseball Reference.

High control (year-over-year r > 0.6): Strikeout rate (K%), walk rate (BB%), and K-BB% (the gap between the two). These are the backbone of FIP and the strongest signal in any starting pitcher projection model.

Moderate control (year-over-year r ≈ 0.4–0.6): Ground ball rate (GB%), line drive rate (LD%), and fly ball rate (FB%). Ground ball rates are more repeatable than fly ball rates for most pitchers, likely because they reflect repertoire and release-point tendencies. High ground ball rates suppress HR vulnerability independent of luck.

Low control / noisy (year-over-year r < 0.3): BABIP on non-HR batted balls, strand rate (LOB%), and HR/FB rate. These are precisely the inputs ERA reflects but DIPS-based estimators deliberately remove or normalize.

Velocity sits upstream of all of these. Statcast data from MLB's publicly available Savant database shows fastball velocity correlates with both swinging strike rate and ground ball tendency, which propagates into K%, BB%, and ultimately FIP. A pitcher losing 1.5–2.0 mph off a fastball mid-season is an early signal worth tracking before ERA moves.

For a broader look at how these statistical inputs connect across projection models, statistical inputs for fantasy projections covers the cross-sport framework.

Classification Boundaries

ERA estimators diverge most sharply in three pitcher archetypes:

Power pitchers with fly ball tendencies: FIP tends to favor these pitchers because it strips out HR variance. xFIP goes further and normalizes HR/FB, which can dramatically lower their xFIP relative to FIP if they pitched in a homer-friendly park or had an elevated HR/FB rate. The gap between FIP and xFIP for a power fly-ball pitcher can exceed 0.50 runs in a single season.

Ground ball specialists: SIERA typically produces the most favorable estimates here. Because ground balls suppress run scoring relative to fly balls — even at similar BABIP rates — FIP slightly undervalues pitchers who induce weak contact at the cost of some strikeout rate. SIERA's interaction terms capture this more precisely.

Soft-tossers with high contact rates: ERA estimators can diverge significantly from each other for this group. A finesse pitcher with a 17% K rate, a 6% BB rate, and a 52% GB rate will score differently across FIP, xFIP, and SIERA because each model weights the contact component differently. This is also where park context matters most.

Tradeoffs and Tensions

The central tension in DIPS-based projection is between purity and predictive completeness. FIP is clean, transparent, and reproducible — but it ignores information that is partially real. Some pitchers do suppress BABIP at rates exceeding random chance over multi-year samples, which suggests either true contact-management skill or consistent defensive support. Knuckleball pitchers and extreme sinkerballers historically show BABIP suppression that persists beyond single-season noise.

SIERA resolves some of this but introduces model complexity that makes it harder to audit. When a SIERA-based projection diverges from a FIP-based projection, it's not always obvious which input is driving the gap.

A second tension: ERA estimators are designed for large samples, typically 150+ innings. For a starting pitcher with 60 innings in April and May, applying full-season regression via xFIP may overweight population-level norms at the expense of real within-season development. This is where sample size and projection reliability becomes directly relevant to SP valuation decisions.

Finally, fantasy scoring formats create their own layer. In formats that score ERA directly (as most head-to-head and rotisserie leagues do), the ERA estimator informs the projection but doesn't replace it. A pitcher who routinely pitches for a poor defensive team will likely allow a higher actual ERA than FIP suggests — and that gap is a feature, not a bug, in projection modeling.

Common Misconceptions

"A lower FIP always means a better fantasy pitcher." FIP is park-neutral and defense-neutral by construction. In formats scoring actual ERA, a pitcher with a 3.20 FIP behind an elite infield defense may outperform a 3.10 FIP pitcher throwing behind one of the league's weakest fielding units.

"BABIP always regresses to .300." The MLB-wide average BABIP tends to cluster near .295–.305, but individual pitchers who consistently induce weak contact — extreme ground ball pitchers, for instance — can maintain below-average BABIP over multi-year windows. McCracken's original finding was about the degree of pitcher control, not that it is zero.

"xFIP is better than FIP for all pitchers." xFIP normalizes HR/FB rate to the league average. For pitchers who consistently induce weak fly ball contact (soft, shallow flies rather than hard-hit bombs), actual HR/FB rate may be sustainably below league average, making xFIP systematically pessimistic.

"ERA estimators predict ERA." They are better described as estimating what ERA would have been under neutral defensive and luck conditions. Actual ERA depends on LOB%, BABIP, and HR/FB rate, all of which carry real variance. The estimators are inputs to projection, not projections themselves.

Checklist or Steps

The following sequence describes how ERA estimators are typically incorporated into a starting pitcher projection workflow:

1. Compare the resulting ERA estimate against rest-of-season projections from at least two independent systems. The comparing projection systems framework explains where systematic divergences tend to appear.

Reference Table or Matrix

The projection models explained overview places these estimators within the broader architecture of how full-season and rest-of-season outputs are built — including where ERA estimators intersect with park adjustments and lineup-level run environment modeling.

For readers building their own evaluation process, what makes a projection accurate examines the empirical tests that distinguish signal from noise across ERA estimator outputs at the individual pitcher level.

The full suite of tools and methodology documentation for starting pitcher analysis is available through the Fantasy Projection Lab home.