Elle's Football Picks
Weekly football picks, scored against reality. No hype, just receipts.
Every fixture from 8 leagues, rated by a stats model. Each card gives you a Safe pick (the model's main call) and a Spicy one (where the bookies look wrong). Picks by league below — running P&L on Bankroll, every result on History.
- Green — Safe pick
- Hot yellow — Spicy with edge
- Amber — Spicy, no edge
- Red % — Bookie has the value
18+ | Gamble responsibly | Odds may change | No guarantee of profit · More info
🇬🇧 Premier League England 10 fixtures
🇬🇧 Championship England 1 fixtures
🇮🇹 Serie A Italy 10 fixtures
Track Record
How does this stack up?
Why ~55% is the ceiling →Football outcome accuracy has a hard ceiling around 55%. Closing-line bookmakers sit at the top of that ceiling — they aggregate every signal and billions in liquidity. The "All" pooled number below is across the top-5 European leagues (the same scope the bookie/academic ceilings use).
Caveat: Per-league numbers are walk-forward eval over roughly 600-1,100 matches per division. Margins under ~2pp are inside statistical noise — La Liga sitting near the bookies isn't a "we beat the market" claim, more that La Liga has historically been one of the more predictable top leagues (consistent top-3 dominance, fewer giant-killings). The bookies' 55% is itself a top-5-leagues pooled figure.
Calibration check
When the model says "65% chance", does it actually happen 65% of the time? Each row shows what the model predicted vs what really happened.
| Predicted | Observed | Sample | |
|---|---|---|---|
| 10% | 0% | 3 | -10pp |
| 17% | 29% | 133 | +12pp |
| 26% | 27% | 797 | +2pp |
| 35% | 38% | 262 | +3pp |
| 45% | 38% | 231 | -7pp |
| 54% | 45% | 152 | -9pp |
| 64% | 52% | 64 | -12pp |
| 73% | 88% | 8 | +15pp |