.jpg)
01Erling Haaland
Manchester City
Independent research / Premier League
Football
Data Lab.
AI-Powered Football Match Analytics
Studying match performance through historical Premier League data, player statistics, tactical indicators and predictive models.
01 / Research foundations
Built on data.
Designed to predict.
Three public-data resources and three modeling approaches—shown together, without unnecessary technical clutter.
Data foundations03 sources
Prediction methods03 approaches
01Demo active
Poisson distribution
Expected goals and outcome probabilities
02Planned
Neural network · ANN / MLP
Nonlinear pattern-learning architecture
03Planned
Large language model
Match context and tactical interpretation
Historical matchesGoals · results · form
Player & tactical featuresPass counts · shots · home/away
Match probabilitiesWin · draw · loss · score
Current website: working Poisson simulation using illustrative coefficients. Historical dataset ingestion, ANN training and LLM inference are not connected to this demo.
02 / League context
The Premier League,
through a data lens.
Different clubs. Different styles. A shared analytical framework.
Team-level indicators
Beyond the
final whistle.
Compare attacking output, defensive records, home advantage and changes in recent form.
Arsenal
Manchester City
Liverpool
Chelsea
Manchester United
Tottenham
Newcastle United
Aston Villa
Brighton
West Ham
Everton
Fulham03 / Player intelligence
Every player adds
another dimension.
From finishing and chance creation to passing patterns: potential inputs for richer match models.
01Goal output
02Bukayo Saka
Arsenal
03Cole Palmer
Chelsea
04Declan Rice
Arsenal
05Mohamed Salah
Liverpool
.jpg)
06Son Heung-min
Player case study
04 / Interactive prediction lab
Choose the match.
Explore the probabilities.
Illustrative Poisson simulation
Adjust club selection and match assumptions. See probabilities, expected goals and possible scorelines.
Home club
Arsenal
Away club
Manchester United
Match outcome probabilities · demo scenario
Match configuration
RUN 01+10%
+10%
Ready to explore scenarios.
Model output
Poisson baselineHome win—
Draw—
Away win—
—Home goal rate (λ)
—Away goal rate (λ)
—Most likely score
Scoreline probability map
0–4 goals per sideRows: home goals · Columns: away goals · Brighter = higher probability
Goal distribution
Poisson probability by countHomeAway
012345+
Research demonstration, not live match forecasting or betting advice. Team strength coefficients are illustrative inputs, not fitted Premier League estimates. Probabilities are computed locally in JavaScript; no trained neural network, LLM or live match feed is connected.
05 / Research process
From raw data
to tested predictions.
A clear workflow. Four steps. Measurable outcomes.
Match evidence
Team contextPlayer features01 Collect
Historical match and player records
02 Engineer
Form, passing and tactical indicators
03 Model
Statistical baseline and AI research
04 Validate
Out-of-sample performance checks
How model performance is evaluated
AccuracyLog lossBrier scoreCalibration