How the gate decides
Most of the time, the honest answer is no.
Everything below was decided by the product's real selective-gate code, run when you loaded this page. The reasons are the gate's own, not copy written to sound careful. It does not yet decide the published board; those refusals still come from a separate stored path, and saying otherwise would overstate where this is wired.
What is real on this page
Illustrative inputsThe decision logic is production code. Nothing here is mocked or written to produce a pleasing answer. The input rows are illustrative: they are not today's slate. Feeding live picks in requires a data join whose behaviour we have not yet verified against real rows, and shipping an unverified join on a page about honesty would be the exact failure this page argues against. Real gate, labelled inputs, never the reverse.
edge threshold τ = 0 · fire when the calibrated lower bound clears the de-vigged price at all, with no added margin
interval = ivap · max width to fire = 0.1 · an interval wider than that means the calibration does not pin the probability down, so the row is refused however good its lower bound looks
What the gate returned
We're on this one. The edge cleared our bar. Public fire held (prefire: live_board_off).
lower-bound edge 0.0156 · interval width 0.0052 · on illustrative inputs
We passed. After the vig, the price didn't give us enough edge.
We're on this one. The edge cleared our bar. Public fire held (prefire: live_board_off).
lower-bound edge 0.0156 · interval width 0.0052 · on illustrative inputs
We passed. After the vig, the price didn't give us enough edge.
We passed. After the vig, the price didn't give us enough edge.
We're on this one. The edge cleared our bar. Public fire held (prefire: live_board_off).
lower-bound edge 0.0156 · interval width 0.0052 · on illustrative inputs
We passed. Not enough history in this category yet to trust our number.
We passed. Not enough history in this category yet to trust our number.
We couldn't score this one. A key piece of data was missing, so this is not a judgment about the game. Missing: q (no two-sided odds captured for this market).
Three different kinds of no
- No bet: we evaluated this against real settled history and the lower bound did not clear the price after vig. A judgement.
- Not judged: this category does not have enough settled history to calibrate against yet, so the model was never asked. Not a judgement, and we will not dress it up as one.
- Not evaluated: an input was missing, so nothing reached the model. Says nothing about the game.
Collapsing these into a single "no bet" would let us claim a considered judgement where the truth is an absence of evidence. For a product this young, the second and third are usually the honest answer, so they get their own names.
What this page does not claim
- No win rate, ROI, or performance result is asserted anywhere on this page.
- A fired decision here is a demonstration of the rule, not a recommendation.
- Nothing here is persisted to the ledger, no receipt is created by loading this page.
- The lower-bound edge printed on a fired row is computed from the illustrative rows above. It is arithmetic on made-up inputs, not a measured edge in any real market.
- The rows are illustrative inputs, not today's published picks.
The governance side of this, how our agents are governed, is a separate claim about a separate subject.