When not to bet
Most published analysis is about which bet to make. The more useful skill is recognising the events where no bet is justified — thin markets, missing information, unpriceable variance, and the moments when the reason to bet is emotional rather than analytical.
Pass when you cannot state, in one sentence, why your probability estimate should be better than the market's. Absent that, the expected value of the bet is negative by the size of the margin. "No strong view" is not a reason to take the favourite — it is a reason to take nothing.
Six conditions that justify a pass
- The price comes from one bookmaker. A single quote is one firm's opinion, not a market. See thin markets.
- Team news is unresolved. If a decisive player's availability is announced an hour before kick-off, the market before the announcement is pricing uncertainty you also cannot resolve.
- The sample behind your view is tiny. Three matches is not a trend. Most "form" arguments rest on samples too small to distinguish signal from noise.
- The market is efficient and you have no private angle. Major-league moneylines are among the best-priced forecasts available anywhere. Beating them requires something the market does not have.
- The variance is unpriceable. Certain markets — first goalscorer, correct score, red cards — carry outcome distributions so wide that a small modelling edge is undetectable inside them.
- You want to bet. Wanting to have action on a match is a reason to watch it, not to stake on it. This is the most common condition and the hardest to admit.
Why is passing the normal outcome?
If the market is right on average and charges 4% to participate, then across all events the expected return of betting at random is −4%. Edges are not distributed evenly: they concentrate in a small subset of events where the market has mispriced something identifiable. Everywhere else, the correct action is to do nothing.
This is where our own engine draws a line that is worth stating plainly, because it is not the obvious one. Agreement between the analytical contours decides whether a signal is published; the price decides only how large it is. A match is passed when the contours point in different directions — not when the price is poor. A poor price produces the minimum volume instead, and the expected value is printed next to the signal even when it is negative. The consequence is that our pass rate is driven by disagreement between models, not by a threshold on expected value. A day with no publishable signals is a normal day, and we publish the pass count rather than hiding it — see why most events get a PASS.
The passes that cost the most
Three patterns produce more losses than any modelling error:
| Pattern | What it looks like | What it actually is |
|---|---|---|
| Chasing | Larger stake after a loss to "get it back" | Position sizing driven by results, not by edge |
| Reverse-engineering | Finding a market to fit an opinion you already hold | The conclusion selecting its own evidence |
| Filling the card | Betting the early game because the late one is hours away | Paying margin for entertainment |
None of these are failures of analysis. They are failures of the decision to analyse at all — which is why they are invisible to anyone reviewing their record by looking at the bets they made rather than the ones they should not have.
Check any of this against our record
Every signal CONSENSUS publishes carries the bookmaker odds fixed before the event starts and the settled result afterwards — including the drawdowns and the losing runs. The running total is on the front page and every entry is in the log.