This is the first entry in our strategy graveyard: a series where we publish the autopsy of every betting strategy we have rigorously tested and buried. Most betting content tells you what might work. The graveyard documents what demonstrably doesn't — with the receipts.
Today's corpse is special, because it is the best-performing strategy we have ever tested. It still lost money.
As part of the investigation that killed our 50 prediction models, we ran a brutally simple simulation: flat stakes on 1,329 bets across 8 European leagues in the 2025-26 season, at real opening odds, with several strategies running side by side on identical matches:
| Strategy | ROI |
|---|---|
| Back every draw | −0.3% |
| Our model + lineup data | −3.1% |
| Blindly back the Pinnacle favourite | −3.6% |
| Our model | −4.3% |
| Always back the outsider | −10.2% |
Read that top line again. A strategy requiring zero thought — put a unit on the X, every match, every league — outperformed months of feature engineering, gradient boosting and expected-goals modelling. It lost only 0.3 units per 100 staked, against a bookmaker margin that should have cost several times that.
The draw has a structural problem: nobody loves it. Fans back their team, casuals back favourites, dreamers back longshots — almost no recreational money wants the 0-0 grind or the late equaliser. Bookmakers know their X liability is small, so the draw price carries less of the margin's weight and drifts closer to — sometimes past — its fair value.
The mirror image is the favourite-longshot bias we measured across ~32,000 matches: longshots are systematically the worst value on the board because everyone overpays for the dream. The draw sits at the opposite pole — the outcome everyone ignores, and therefore the one the market prices most honestly.
That's also why "back every outsider" lost 34 times more than "back every draw" in the same simulation. Same matches, same bookmakers — the only difference is which psychological bias you're standing on.
No — and this is where most betting content would start lying to you. −0.3% is still a loss. Run it forever and you fund the bookmaker, just very slowly. What it is: proof that the draw is where the margin is thinnest, which makes it the only 1X2 corner where a selective approach might plausibly clear zero.
Blanket-backing every draw fails because it includes matches where the draw is honestly priced or worse. The interesting question is whether some identifiable subset of draws is systematically underpriced.
While digging through six seasons of Italian football data, we found something that made us sit up: draws in Serie A and especially Serie B appear systematically undervalued — in Serie B the effect showed roughly +7% expected value on draws at odds up to 3.5, positive in 6 out of 6 seasons, with a confidence interval that excludes zero.
We know exactly how stories like this usually end — we wrote the article about lucky backtests ourselves. One league, one pattern, found after looking at the data: that is precisely the shape of an artefact. So we did the only honest thing available: we pre-registered the strategy — rules fixed in advance, hashed and anchored in Bitcoin, before the season — and starting August 2026 we are running it live in our forward test, every pick logged before kick-off and settled against the closing line, in the public track record.
If the Italian draw effect is real, you will watch it prove itself in public. If it's a mirage, you will watch it die in public — and it will get its own entry in this graveyard, right next to the strategies it outlived.
Every claim above comes from our own measured data — the simulation details are in the full investigation write-up, the draw test goes live in August, and the numbers will land in the track record whether they flatter us or not.