Our approach
Spectral uses statistical regression models calibrated per league to predict total corners in football matches. By comparing model predictions against bookmaker odds, we identify value bets where the true probability exceeds the implied odds. Every prediction is timestamped before kick-off, and every settled pick — free or paid — appears on the public track record.
The model
Statistical regression
Purpose-built statistical regression models, not off-the-shelf classifiers. Designed specifically for predicting corner totals across different leagues and contexts.
49 leagues worldwide
Trained on historical match data across 49 leagues. The breadth of coverage means the model can surface value in markets most tipsters never look at.
Per-league calibration
Every league plays differently. Per-league calibration accounts for different playing styles, corner frequencies, and market dynamics so predictions stay accurate across competitions.
Feature engineering
Team form, shooting stats, possession patterns, and league position all feed into predictions. The features are selected based on what actually drives corner counts, not what sounds impressive.
Confidence scoring
Every pick gets a confidence tier based on model fit, data availability, and calibration quality. High-confidence picks come from leagues where the model has strong historical accuracy and deep data coverage. Low-confidence picks are flagged honestly so you can size your stakes accordingly.
Finding value
Probability estimates
The model produces probability estimates for corner totals at each line. These are genuine probability distributions, not just point predictions.
Odds comparison
Model probabilities are compared against bookmaker odds to find genuine edges. If the model says over 9.5 corners has a 55% chance but the bookmaker implies 48%, that gap is the edge.
Threshold filtering
Only picks where the edge exceeds confidence-adjusted thresholds make the cut. The threshold is higher for low-confidence leagues and lower for leagues where the model has proven accuracy.
Kelly stake sizing
The Kelly criterion determines optimal stake sizing based on the size of the edge and the confidence level. This maximises long-term bankroll growth while managing risk.
Transparency
Pre-kickoff timestamping
Every pick — free or paid — is timestamped before kick-off. Free picks publish on the site; paid picks go to subscribers in the dashboard and via Telegram. After settlement, every pick of every tier is counted in the track record. Nothing is edited after the event.
Calibration curves
Calibration curves show how well predictions match reality. If the model says 60% probability, does it actually happen 60% of the time? We publish the answer.
Model performance metrics
Brier score and skill score are published for every league. These measure how well the model predicts, compared to simply guessing the base rate.
Full P&L breakdown
Complete profit and loss tracking with league-by-league breakdowns available on the track record page. The good, the bad, and the ugly.
How we measure performance
Every pick is timestamped at real bookmaker odds and tracked through to settlement. Realised yield is measured over rolling windows, broken down by league and by month. The full history lives on /track-record.
Spectral focuses on corners markets across 49 leagues. Corners get less modelling attention from bookmakers than main markets, which is where the edge comes from. The trade-off is that the closing line on corners isn't always a sharp benchmark, so we don't pitch “we beat the bookmakers’ closing line” as our credibility metric. Performance is what's on the track-record page, not a closing-line beat rate.
Behind the scenes, we snapshot the closing line where it's available (5 to 20 minutes before kick-off, across five UK and exchange books: Bet365, Betfair Exchange, 888, Coral, Paddy Power). Pinnacle isn't in the panel. On corners markets specifically the closing line is noisier than on main markets, so we treat it as an internal diagnostic rather than a public credibility claim. Realised yield is what we're willing to be measured on.
Why corners?
Corners is a niche market where bookmakers are less sharp. The major sportsbooks focus their sharpest traders on match result, over/under goals, and Asian handicap markets. Corners gets less attention, which means the odds are softer and edges persist for longer.
Less attention from sharp bettors means more opportunity for statistical models. The corners market is less efficient, and the edges tend to be wider and more consistent than in heavily traded markets.
Rich historical data makes corners particularly well-suited to quantitative analysis. Corner counts correlate with measurable match features like possession, shots, and attacking style, which makes them predictable with the right model.
As the model matures, we'll expand into additional markets where we can identify a genuine statistical edge.