What your target and stop are worth before any read. Priced off a validated 192-cell race model with the deadline exit modelled — so it answers the whole question, not just who wins when somebody wins.
| round trip | 0.25 pt | 0.50 | 1.00 | 1.70 | 2.50 | 4.00 |
|---|---|---|---|---|---|---|
| EV, R |
Cost is charged against your stop distance, so a tight stop pays the same ticks over a smaller denominator. That is the whole tax: 0.11R at a 4.5-point stop against 0.007R at 68 points — a 15× difference before any edge exists.
Rows are target distance, columns stop distance, both as a fraction of the forecast range. The diagonal is exactly zero before costs — equal distances are a coin flip, which is the check that the model is sound. Everything off it is real market structure.
What this is, precisely. A 192-cell table of P(target first), P(stop first), P(neither) and the average R of the bell exit, measured over 57,251 unconditional entries across 734 NQ sessions, every race resolved on the tick tape so ordering is exact. It reproduces the desk's verified stop-touch surface to within 1.3–2.6 pp and its equal-distance diagonal lands on exactly 50.0%.
What it is not. It has no opinion on direction and no knowledge of your read. It answers one question: if I have no edge at all, what does this geometry cost me? If it says −0.15R, your read has to be worth more than +0.15R before you make a cent. That is the number to know before you click, not after.
The honest limit. Between 0% and 63% of races never resolve inside the hold, and the bell exit is modelled as an average, not a distribution — so treat the EV as a centre, not a guarantee. Fitted on 2022-01 → 2024-12 only; 2025 onward is deliberately unopened and stays that way.