Why most events get a PASS
CONSENSUS runs a multi-layer analytical process over the same event and publishes only where the independent checks agree. Most events end in disagreement and produce nothing. Why that filter exists, what it costs, and why volume is the wrong product metric.
CONSENSUS publishes a signal only when its independent checks converge on the same market. Most events do not produce convergence, so most events produce nothing. The filter is the product: a short list of signals where independent analyses agreed is worth more than a daily quota of signals where they did not.
What does independence mean between analytical contours?
The same event is analysed in parallel along independent paths, and no path sees another's conclusion. Each works from the same underlying data with a different analytical profile. Only after all have finished does the engine compare their outputs.
Independence is what makes agreement meaningful. If models saw each other's work, convergence would measure influence rather than evidence. This is the same reason a panel of forecasters is polled separately before discussion.
The three outcomes
| Status | Condition | What happens |
|---|---|---|
| CORE | All three analyses converge on the same market | Published at full volume |
| RISK | Two converge and the third is neutral, or the estimates are spread wide | Published, labelled as lower conviction — never treated as equal to CORE |
| PASS | The analyses point in different directions, or none of them converge at all | Logged, nothing sent. This is the most common outcome |
A direct contradiction between two analyses is treated as a stronger reason to pass than simple absence of signal. Two well-argued opposite conclusions on the same market means the event is genuinely uncertain — and a market that is genuinely uncertain is priced approximately correctly.
What the price does and does not decide. Notice what is absent from the table: the price. Convergence decides whether a signal is published; the divergence between our estimate and the market decides the size of the position and nothing else. A price that works against us produces the minimum volume, not a pass — and the expected value is printed next to the signal, negative sign included. The reasoning is that a pass is a claim about the event being unreadable, and a poor price is not evidence that an event is unreadable. Mixing the two would let a pricing decision masquerade as an analytical one.
Why is volume the wrong metric?
A prediction service optimising for published volume has to find something to say about every fixture. Most fixtures are efficiently priced, so forcing output on them means publishing signals with no edge — which drags the aggregate towards the break-even rate and adds margin costs on top.
The arithmetic is unforgiving. Adding coin-flip signals to a set of good ones does not dilute the edge proportionally; it destroys it, because each added signal pays the bookmaker margin. A service that publishes ten signals a day is telling you it has found ten edges a day, which is a much larger claim than it sounds.
This is also why the number of published signals will never be a target here. A quota converts a filter into a content schedule, and the filter is the only thing worth paying for.
What does the filter cost?
Honesty requires stating the downside. A strict agreement rule reduces sample size, and a smaller sample takes longer to demonstrate anything statistically. It also means some genuinely good opportunities are passed over because one analysis was cautious. The filter trades coverage for precision, and that trade is a choice rather than a free win.
It also makes the record slower to build. A service publishing thirty signals a week reaches a thousand settled results in eight months; at our rate it takes considerably longer. We accept that, because a thousand results from a forced quota would not answer the question a hundred honest ones do.
How do you read a PASS?
A PASS is not a prediction that nothing will happen, and it is not a hedge. It means the method did not produce a scenario stable enough to publish, and that the event stays in the log without a signal.
From the reader's side, the value of a PASS is the events it removes from consideration. You are paying for the events we talk you out of as much as the ones we talk you into — and unlike a published signal, the value of a skipped event never appears in any performance statistic.
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.