Football analytics and match modelling

Math & Bets · Football Weekly Ratings

The Global Football GLM Engine.

Weekly football ratings built to turn model outputs into prices you can compare with the market.

A football forecasting engine built around a Quasi-Poisson Generalized Linear Model, Expected Goals, time-decay weighting and Dixon-Coles low-score adjustments — published as a practical weekly ratings matrix.

Model estimates are informational only and do not guarantee future results.

The engine

Look past the final score.

Raw goals can be noisy. They are affected by finishing variance, game state and individual events that may not reflect the underlying flow of a match.

Math & Bets evaluates teams using Expected Goals generation and suppression when suitable xG data is available, then converts those inputs into match and handicap probabilities.

What you get every week

A clean matrix you can actually use.

Updated weekly
01

Weekly ratings file

A single spreadsheet with the selected fixtures, projected xG, cover probabilities and model-implied prices.

02

Asian Handicap matrix

Quarter- and half-line probabilities across the board, so you can compare the model with the price available in your market.

03

Price thresholds

Clear model-derived reference points that show where the market price begins to differ meaningfully from the model estimate.

04

League-by-league coverage

The slate changes with the football calendar and available data, with the model adapting to the quality of each competition.

Current weekly release

Football Weekly Ratings

Delivered through Gumroad. One place to access the current release and future updates attached to the product.

Access this week’s ratings

Under the hood

Three modelling layers.

01

Exponential Time Decay

Recent matches receive more weight than older matches. The current configuration uses a 120-day half-life, so stale observations fade naturally rather than disappearing abruptly.

02

Dixon-Coles Correction

Low-scoring football outcomes are not perfectly independent. A Dixon-Coles adjustment is used to refine probabilities around scorelines such as 0-0, 1-0, 0-1 and 1-1.

03

Multi-League Inputs

Where reliable xG data exists, the engine uses it. For competitions without consistent xG coverage, the model can fall back to goal data with more conservative variance handling.

Football Weekly Ratings

The Asian Handicap Matrix.

Every week, the model produces match-level estimates across quarter- and half-goal Asian Handicap lines. The output includes cover probabilities, model-implied prices, projected xG and price thresholds for each listed market.

Cover probabilitiesModel probability by handicap line.
Fair / implied oddsA direct comparison point against market prices.
Projected xGExpected scoring output for each side.
Price thresholdsReference prices for comparing model output with the market.
Math & Bets weekly football ratings spreadsheet
Example weekly ratings output. Actual fixtures and model outputs change each week.

Get the current release

Stop guessing. Compare the market to the model.

Access the weekly ratings

Workflow

From match data to a usable price.

1Match dataxG or goals, depending on league coverage
→
2GLM engineTime weighting + team attack/defence effects
→
3Score probabilitiesIncluding Dixon-Coles low-score adjustment
→
4Handicap matrixCover probabilities and model-implied odds

FAQ

What the ratings are — and aren’t.

What exactly do I receive?

You receive the current Football Weekly Ratings release through Gumroad, presented as a spreadsheet-style ratings matrix with projected xG, model probabilities, implied prices and market-comparison thresholds.

Which leagues are covered?

Coverage varies by week and by data quality. Competitions with reliable xG inputs are preferred; lower-data leagues may use actual-goal inputs with more conservative handling of uncertainty.

When is the file updated?

The product is designed as a weekly release. Exact publication timing can vary with the fixture calendar and data availability.

Are these betting tips?

No. The ratings are model estimates designed to help compare a mathematical price with a market price. They are not guarantees of future outcomes.

Does the model always use xG?

Where suitable xG data is available, it is preferred. For leagues without consistent xG coverage, the engine can use actual-goal inputs with more conservative treatment of uncertainty.

What is the Dixon-Coles adjustment?

It is a correction commonly used in football score models to improve the treatment of dependence in low-scoring outcomes.

How often are ratings produced?

The product is designed as a weekly ratings release. The exact competitions and fixtures may vary with the football calendar and available data.

Where is payment and delivery handled?

Purchases and digital delivery are handled through Math & Bets on Gumroad.

Math & Bets

From intuition to probability.

Use the model as a second opinion — then decide for yourself.

Access this week’s ratings