Four-model Sports methodology

October 15, 2026 · Four-model Sports edition

Classic, Primary, Tighten Up Conservative ON and OFF are shown separately. The published selection board uses Tighten Up OFF. ON/OFF share scoring arithmetic; four configurations contain three score engines. NBA and NHL remain tracking-only. No model probability is calibrated.

Compare all four models · Methods and limitations · Results tracker

One edition, four traceable configurations

Classic 2.0 and Primary 2.0 are explicitly new provisional engines. Tighten Up ON and OFF use the same registered score equations and different eligibility gates. The OFF board continues the latest approved website configuration, with all four shown in each story and comparison. No model averaging creates a fifth forecast.

Replacement and trade review

Every covered football role is reviewed; qualifying availability scenarios retain the separately authorized −3 sensitivity. It is not a fitted player coefficient. Base, adjusted and joint-overlap scenarios remain in the immutable record. Trades receive separate sending/acquiring offensive and defensive reviews without automatic −3. MLB/NBA/NHL offensive comparisons preserve role opportunities and team-split histories. NBA/NHL are tracking-only. Unknown effects remain unestimated.

Eligibility and invalidation

OFF football floors are 3 points on spread and 4 on total; ON floors are 4 and 5. Every selected case must survive the prescribed 3-point adverse test, a total retaining at least one point, and applicable opponent-offense and replacement branches. High risks force PASS; two medium risks require downgrade and three force PASS. ON has at most two tiers and no Free Picks; OFF has at most two Free Picks. MLB OFF run-line and total floors are one run; ON requires1.5. The declared one-run adverse case is applied once and event-specific personnel pathways are reviewed separately, without double-counting the same workload concern. Three unresolved medium risks prohibit a tier; two require at least one downgrade. Every recommendation must be strictly better than −400. Model agreement is not a calibrated probability.

Current edition

No prior edition-only exception carries forward. Fresh evidence and score attachments were frozen before fresh prices. Conditional player effects remain distinct from measured adjustments. The NFL forecast uses the authorized 37.5% prior and 62.5% current actual-drive blend; descriptive error allowances do not establish calibrated probability. Every current comparison preserves the registered formulas and separately labeled player scenarios.

Tracker and grading

Every published event has one tracker row with all four configurations attached. Any incomplete-market exclusion remains explicit in the current coverage audit. Forecast grades are separate from wager results. The verified October 14 OFF results are retained in the tracker; the inherited historical record is not OFF validation. Weighted historical units are not cash profit. Missing confidence and stakes do not become zero-probability forecasts or zero-dollar wagers. New prospective scores await official finals; elapsed coverage has no forecast to grade.

Audit and sources

Forecasts are informational, not guarantees or wagering advice. Use licensed operators, set limits, never chase losses, and stop when betting is no longer entertainment. No guaranteed return, calibrated wagering probability or assigned stake is claimed.