Methodology
Practical implementations and approachable write-ups covering data, models, and market concepts.
Getting Started
Foundational concepts for football analytics and betting
Intro to Football Data
Data6 min read
What football data looks like, common sources, and how to ingest it for analysis.
Read ArticleIntro to Football Betting
Betting6 min read
Foundations of market odds, value, and basic risk management for betting experiments.
Read ArticleOur Models
How we built and iterated on our own prediction models
March Madness: Spreads Beat ML
Case Study12 min read
How 15 iterations taught us that Vegas spread lines crush hand-crafted features — our journey from box-score ML to a spread-first Elo-nudge model scoring 0.132 Brier.
Read ArticlePremier League: Poisson + Dixon-Coles
Case Study15 min read
Building our own match odds using rolling xG averages, ClubElo ratings, and a Poisson goal model with Dixon-Coles correction — from data pipeline to live UI.
Read ArticleBacktest Dashboard
ResultsDashboard
Interactive charts showing cumulative P/L, calibration, value bet decisions, and match-by-match results from our walk-forward backtest on 200+ EPL matches.
Read ArticleBetting
Market mechanics, pricing, and risk
Arbitrage Betting
Betting8 min read
How to detect pricing inconsistencies across bookmakers using implied probabilities, and why execution risk matters.
Read ArticlePrediction Markets
Markets12 min read
How Polymarket, Kalshi, and betting exchanges work — trading binary contracts, order books, and applications for sports.
Read ArticlePredictive Models
Machine learning algorithms for match outcome prediction
Regression Models
Fundamentals18 min read
From linear to logistic to Poisson: understanding distributions and choosing the right regression for your target variable.
Read ArticleRandom Forest
Ensemble Methods20 min read
The wisdom of crowds: combining many decision trees through bagging and random feature selection for robust predictions.
Read ArticleGradient Boosted Machines
Ensemble Methods20 min read
From decision trees to XGBoost: understanding sequential ensemble methods and their application to match prediction.
Read ArticleKalman Filters
State Estimation30 min read
How to estimate truth from noisy measurements — the maths behind smooth player tracking, velocity estimation, and sensor fusion in football analytics.
Read ArticleBayesian Modelling
Probabilistic Inference35 min read
How to reason under uncertainty — updating beliefs with evidence, quantifying what you don't know, and making better decisions when data is scarce.
Read ArticleMonte Carlo Simulations
Computational Methods30 min read
When you can't solve it analytically, simulate it 10,000 times and count — from season prediction to bankroll management.
Read ArticleMixed Effects Models
Statistical Modelling30 min read
When your data has structure — players nested in teams, matches nested in seasons — partial pooling gives correct answers where standard regression fails.
Read ArticleGeneralised Linear Models
Statistical Modelling30 min read
The framework that unifies linear regression, logistic regression, and Poisson regression — from xG models to match prediction.
Read ArticleTime Series Forecasting
Predictive Models30 min read
Predicting what happens next using the structure of time — from Elo ratings to season-long xG trajectories and injury risk curves.
Read Article+
More models coming soon
Neural Nets, Bayesian Models, etc.
Spatiotemporal Graph Neural Networks
Deep learning architectures for modeling player interactions and match dynamics
1. Intro to Neural Networks
Deep Learning45 min read
A comprehensive beginner-friendly introduction to neural networks — from single neurons to deep architectures, with intuitive explanations and football examples.
Read Article2. Convolutional Neural Networks
Computer Vision40 min read
How neural networks learn to 'see' — from detecting edges to understanding complex patterns in images, videos, and football pitch data.
Read Article3. Recurrent Neural Networks
Sequence Modeling45 min read
How neural networks learn from sequences — remembering the past to understand the present, from player trajectories to match narratives.
Read Article4. Graph Neural Networks
Graph Learning50 min read
How neural networks learn from relationships — modeling player interactions, passing networks, and the complex web of football dynamics.
Read Article5. Spatiotemporal GNNs
Spatiotemporal55 min read
The ultimate architecture for football analytics — combining GNNs with temporal modeling to understand how player interactions evolve during a match.
Read Article6. Attention & Transformers
Attention45 min read
The revolutionary mechanism that lets neural networks focus on what matters — dynamic weighting of players, positions, and moments in football analytics.
Read Article7. Systematic Literature Review
Academic Review35 min read
A comprehensive academic review of STGNN applications in sports analytics — surveying foundational architectures, current research, open challenges, and future directions.
Read ArticleSTGNN Implementations
Real-world case studies and practical implementations of spatiotemporal models
SkillCorner Analytics Cup: STGNN Approach
Case Study30 min read
A deep dive into Zach Cochran's winning approach to the SkillCorner tracking data competition — implementing STGNNs for player trajectory prediction.
Read ArticleTranSPORTmer: Unified Sports Trajectory Modeling
Research Deep Dive35 min read
How TranSPORTmer uses set attention to handle trajectory prediction, imputation, and classification in a single unified architecture.
Read ArticleDiffoot: Graph-Conditioned Diffusion Models
Research Deep Dive40 min read
How diffusion models combined with graph neural networks generate diverse, multi-modal trajectory predictions for tactical analysis and counterfactuals.
Read ArticleCausalTraj: Coherent Multi-Agent Forecasting
Research Deep Dive40 min read
How temporal causality and joint likelihood modeling generate coherent multi-agent trajectories — where players move together realistically, not just individually accurately.
Read ArticleSportsNGEN: Sustained Gameplay Simulation
Research Deep Dive35 min read
A transformer-based simulation engine that generates realistic, sustained gameplay — capable of simulating entire tennis matches and football sequences with counterfactual analysis.
Read Article+
More implementations coming soon
Custom STGNN builds & experiments
Miscellaneous
Engineering notes and general tooling
Entity Matching
Data Engineering10 min read
Solving the challenge of linking players and teams across data sources with fuzzy matching, embeddings, and metadata.
Read ArticleVideo Analysis Pipeline
Computer Vision15 min read
How we turn broadcast footage into pitch coordinates — detection, tracking, calibration, and projection explained step by step. Plus how the industry does it better.
Read Article+
More articles coming soon
Data pipelines, tooling & utilities
Implementation Overview
How we apply theoretical concepts to real match data
The methodology section focuses on practical, reproducible approaches. Each article explains a specific topic, shows how data is collected and cleaned, and outlines simple code or SQL patterns to reproduce the steps. Articles are intentionally concise and actionable.