303,000+ matches · 26 years of closing odds · 0findings survived out-of-sample validation · every number after the bookmaker's cut

Is the betting market wrong?

We replayed history: what if you had blindly made the same type of bet on every match since 2000? Each result below is that experiment — profit or loss per 100 staked, at real closing odds, margin included. This is the foundation the prediction model gets built on.

The longshot tax

—

blind-backing under-10% shots · undefined bets. The crowd overpays for miracles.

Favorites are honest

—

backing 90%+ favorites nearly beats the margin (small sample: undefined bets)

Draws myth: busted

0.0%

blind-backing every draw, 0 bets — draws are NOT underpriced

Laziest market found

—

The famous curve: what blind betting returns, by market confidence

760k selections · green above zero = profit

Read it left to right: the less likely the outcome, the worse blind-betting it performs — from −30 per 100 on longshots to roughly break-even on heavy favorites. This is the favorite-longshot bias, the most replicated result in betting economics — and our pipeline reproducing it from raw data is the proof the measurement machinery works. Where the crowd systematically overpays, the other side of those bets is systematically cheap.

Full findings register

1 findings · timestamped · misses shown like hits

n = number of bets in the replay · ROI = profit per 100 staked · 95% CI= the honesty range (if it includes 0, luck can't be ruled out) · validated = confirmed on matches the test never saw, at market-average prices, under two different margin models · candidate = real beyond luck on pooled history, not yet enough recent data to confirm · finding only = not distinguishable from chance.

Show all 1 findings

Finding	n	ROI %	95% CI	Status
Blind-backing every draw	0	0.00	0.0 … 0.0	finding only