How It Works
We use advanced statistics to measure the gap between our estimate and what the odds imply.
Every day, bookmakers set odds on thousands of football matches. Sometimes their implied probabilities differ from what our model estimates. We use a mathematical model to measure that gap — and show it to you.
The Process
1. We Collect the Data
Every day before matches kick off, we pull live team statistics, form, head-to-head history, and bookmaker odds from the leagues we cover worldwide. We track 50+ metrics per team: goals scored and conceded, win rates, clean sheets, BTTS rates, corner stats, and more.
2. We Calculate True Probabilities
Using our statistical model, we estimate the real probability of each outcome — Home Win, Draw, Away Win, Over/Under 2.5 goals, Both Teams to Score, and more. These are what the odds should be, based purely on data.
3. We Compare With the Bookmakers
We remove the bookmaker's margin from their odds, then compare their true implied probability with our model's estimate. When our probability is significantly higher, we publish the gap as a selection — what you do with it is up to you.
What Makes Us Different
- ✓We don't guess. Every pick comes from a mathematical model run fresh each day.
- ✓We don't follow tipsters. Our model is fully independent — no human bias.
- ✓We don't cherry-pick. Every analyzed match is included in our track record.
- ✓We calibrate per league. Each league we cover has individually optimized parameters — what works in the Premier League may not work in Serie A.
Think in months, not days
Even a strong model gets ~4 picks out of 10 wrong. Bad weeks happen — and they're mathematically expected. A bettor following picks for 30 days will see variance. Over 3-6 months, results stabilize. Judge any model over 100+ bets, not 10. That's how professional bettors think.
Our Track Record — Live
Published monthly from launch, win or lose: every pick, every result, with sample sizes shown. No cherry-picking — judge for yourself.
For the Technically Curious▾
Our probability engine uses a Poisson distribution with Dixon-Coles low-score correction (1997). Attack and defense strength vectors are computed from season stats blended with last-5 form (dynamic blend: 55-80% Poisson / 20-30% rolling form / 0-20% H2H — weights adjust based on data availability). Probabilities are dampened using a conservative blend between model estimate and market-implied fair odds. The gap calculation uses multiplicative devigging to remove the overround; only selections with a positive gap to market pass the filter. Each selection passes multiple filters: minimum odds threshold, minimum gap, confidence tier (HIGH/MEDIUM), and per-league market blacklists.