EPL Projection Model

What goes into a team's expected goals?

Five components: the goals a team actually scores and concedes at each venue, how well it suppresses opponents, its recent form, fixture congestion, and home venue. A sixth factor, player absence, then cuts the number when the announced starting eleven is missing expected starters.

The model computes each team's scoring rate (lambda) from five weighted components. Goals differential (each team's actual home/away goal splits) is the foundation; a 2025-26 audit found the model's prior xG-based version produced systematic per-team biases, so live lambdas are built from actual goals instead. The remaining components capture factors those splits alone miss. The percentages shown describe the design allocation across components; the live formula applies each factor as a multiplicative adjustment rather than consuming these percentages directly.

What exactly does the tau correction change?

Four cells of the scoreline matrix, and nothing else. It raises 0-0 and 1-1, lowers 1-0 and 0-1, and leaves every other scoreline exactly where plain Poisson put it. One fitted parameter, rho, sets how far each of the four moves.

How much does recent form move the number?

It shifts a team's expected goals by a fixed percentage for each goal per match its recent scoring differs from its season rate, and the shift is capped in both directions. Every match in the window counts the same. The model applies no time decay inside the window.

Football teams evolve through transfers, tactical adjustments, injuries, and manager changes. The live model takes an arithmetic mean over the trailing finished-fixture window, so every match in that window carries equal weight. It does not apply exponential time decay.

How does the model handle home advantage?

Each club's home and away scoring and conceding rates are estimated separately, so a fortress ground is already in the expected goals before any further adjustment. A 2026-07 test added a generic home boost on top and found it bought no 1X2 accuracy while making goal totals worse, so no extra boost ships.

Home advantage in the EPL is carried by the per-team venue-split strengths above: each club's home and away scoring and conceding rates are estimated separately, so a fortress ground or a difficult away trip is already priced into the lambda before any further adjustment. A 2026-07 preregistered ablation tested adding a generic multiplicative boost and per-ground premiums (Anfield, St James' Park) on top of the venue splits by replaying three completed seasons through the real backtest. The extra boost added no measurable 1X2 accuracy and made goal-total projections measurably worse, so both the boost and the premiums are neutral in the live model.

What do fixture congestion and a new manager do to the projection?

Congestion cuts a team's expected goals when it plays again before the rest threshold below, because European and cup fixtures force rotation and tired legs. A new manager raises them by a fixed percentage that decays to zero over his first matches.

These adjustments capture situations that baseline team strength doesn't account for. Fixture congestion from European competition forces rotation and fatigue. A new manager appointment produces a well-documented short-term performance boost as players respond to fresh methods and increased scrutiny.

What league averages anchor the numbers?

Home and away goal averages are held separately, because a home team and an away team score at different rates across the league. Every attack and defense rating is measured against the matching baseline, and they refresh each matchweek.

The league-wide scoring averages that anchor all lambda calculations. Home and away baselines are separated because home advantage in football is baked into the attack/defense strength ratios, with home teams creating better chances on average.

What happens when the model and the sportsbook disagree?

Fairline writes an internal observation row and shows you nothing new. A row is recorded when the model's fair odds and a book price differ by a flat 3%, and those rows feed calibration and closing-line research. They are never presented to you as a bet to place.

Fairline records an internal observation whenever the model's fair odds differ from the sportsbook's price by a flat 3%, the same threshold for every market. Each row carries a one-unit research weight for calibration and closing-line tracking. These observations are not surfaced as bets.

Command-line reference. The standalone model runner also includes a quarter-Kelly staking calculator with per-market EV thresholds (below). These are an offline reference and do not drive any user-facing surface. The internal measurement record uses the flat 3% threshold and one-unit research weight described above.

Which Premier League markets does the model price?

Six: the 1X2 match result, the Asian handicap, goal totals, both teams to score, correct score, and team totals. Every one is read off the same Dixon-Coles scoreline matrix, so a correct score price can never contradict the 1X2 price built from the same cells.

Football has more distinct market types than any other sport the model covers. Each one asks a different question, and the Dixon-Coles score matrix lets the model answer all of them from a single unified probability distribution.

Has the EPL model been validated against the market?

Not in a forward, preregistered comparison against a sharp closing line. A 380-match season is too small for accuracy figures to mean much on their own, so the model is checked on Brier score and closing line value instead, graded as matches settle.

EPL presents an honest backtesting challenge because the season is only 380 matches, a sample too small for accuracy metrics to be meaningful on their own. Dixon-Coles is validated primarily through Brier score and CLV. The core question is whether the model's probabilities are better calibrated than the market's opening lines. Historical academic replications of Dixon-Coles on multi-decade European football data consistently show it is well-calibrated against match outcomes, and our implementation follows the same framework. Live validation happens continuously against bet results on the Ledger tab on Transparency.

FairLine has not run a forward, preregistered comparison of the EPL model against a sharp closing line; that needs more forward finals than one 380-match season provides (T7.8). Status: pending data.

When should you not trust this model?

Trust it least on midweek matches after a Champions League tie, on the first match back from an international break, and in matchweeks 1 to 3. Trust it most on a standard Saturday fixture with a full week of rest, from matchweek 6 on. The two lists below give the full version.

EPL's weekly rhythm means most matches are clean projection environments. The exceptions are European-competition weeks and international breaks, where fixture context fights the model's assumptions.

Where can I read the draw correction in more detail?

The tau correction is the part of this model people ask about most. Why do soccer models adjust the odds of 0-0 and 1-1 draws works all four corrected cells through one match, with the arithmetic shown.

The parameter values on this page are rendered from the running system and refresh periodically; when a weight or threshold changes, this page reflects it automatically.