4DFOURTH DOWN LABSJoin the huddle

FREE TOOL / FDL-001 MODEL

Expected
Touchdown Lab.

Put a receiving season through the exact opportunity baseline used in our published touchdown-regression study. See the estimate, the gap, and every coefficient—without sending us your inputs.

MODEL STATUSHISTORICAL BASELINE

Not a 2026 projection

TRAINING WINDOW2014–2023

1,336 player-seasons

STUDY POPULATIONWR / TE · 40+ TGT

Inputs below threshold are flagged

INPUT COLLECTIONNONE

Browser-only calculation

SYNTHETIC STARTING POINTS

PRIVATE BY ARCHITECTURE. This calculation runs in your browser. There is no submit button, account, input storage, or server request.

EXPECTED TD7.27

Opportunity baseline

ACTUAL TD10

Your input

TD − xTD+2.73

Actual minus baseline

INTERPRETATION

Scoring finished above the opportunity baseline.

Use the gap as a weak historical tiebreaker. Do not subtract it from a projection. Preserve opportunity first, then update quarterback, team, health, red-zone role, and market context.

OPEN FORMULA

Correlated inputs make individual coefficients non-causal.

Intercept0.330
Targets 130 × -0.03326-4.324
Receptions 84 × 0.01193+1.002
Receiving yards 1,180 × 0.00196+2.312
Receiving air yards 1,480 × 0.00126+1.860
Receiving first downs 54 × 0.11274+6.088
RAW SUM7.2680LOWER CLIP0

HOW TO USE IT

Regress conversion.
Preserve opportunity.

Use touchdown-over-expected only as a weak historical tiebreaker after preserving opportunity and updating the current team, quarterback, health, and market context.

Read the complete study →

WHAT THE MODEL IS

ridge-regularized linear expected-touchdown model

Published run
2026-08-30-touchdown-regression-v1
Test season
2024
Ridge penalty
5.0
Frozen validations
3 / 3
Parameter SHA
ebd6cd8d2895…
Open model parameters ↗Open tool contract ↗

LIMITS / VISIBLE BY DEFAULT

A calculator is not
a projection system.

  1. This is the published 2024 walk-forward baseline trained on 2014–2023 seasons, not a 2026 player projection.
  2. Inputs below 40 targets are outside the study population.
  3. The five opportunity features are correlated; individual coefficients are not causal effects and should not be interpreted in isolation.
  4. The model omits route-level, red-zone, quarterback, team-environment, health, and betting-market information.
  5. Next-season validation conditions on players returning to at least 40 targets, creating survivor bias.
  6. Expected touchdowns were modestly better than raw touchdowns as a population baseline; both forecasts still missed by more than two touchdowns on average.
AI DISCLOSURE

GLM-5.3 did not calculate these coefficients and is not called when you use this tool. Deterministic Python fit the published model; the frozen launch study used deterministic review checklists and was not GLM-5.3-reviewed. The current GLM-5.3 workflow and its failures remain public in the AI operations ledger →