Why the biggest 'edges' in sports betting are usually wrong
Across football and baseball, our graded results show the same pattern: when a model disagrees hugely with the market, the model is usually the one that's wrong.
Imagine your model says a team has a 60% chance and the market prices them at 40%. Twenty points of value! The honest question: which is more likely — that thousands of bettors with money at stake all missed something, or that your model doesn't understand this game?
What the data says
We graded it. In our football record, every early loss sat on a long price with a big claimed edge, while everything at short prices won. In baseball, claimed edges above 12 points hit 33% with a −43% return — while every moderate-edge band was profitable. Two sports, one lesson: huge disagreement with a liquid market is a bug report about your own model.
The Kelly trap
It gets worse, because proper bet sizing scales stakes with claimed edge. If your biggest errors produce your biggest claimed edges, Kelly obediently puts the most money on your worst reads. Without a cap, a model's blind spots become its largest positions.
What discipline looks like
So the engine refuses its own best-looking bets: hard ceilings on price, hard caps on claimed edge, and every refusal published with the reason. It costs the occasional miracle winner — and buys a strike rate. On the MLB board, a huge model-market gap is flagged as interesting, never as free money. Read it as a question, not an answer.
Frequently asked questions
What counts as a suspicious edge?+
In our graded data, claimed edges beyond about 12 percentage points against a liquid market lost heavily, while modest edges of 3-10 points were profitable. A liquid market is rarely wrong by a lot — a model often is.
Every fixture, fully modelled — the correct-score grid, the derived markets, and the written read.
Start free