Where the windows are (§10.3, measured)
loading…
| Decile of shoe | Fire rate |
|---|
Detector A — window-hit-rate test
Needs the pack read. Count the player's pair bets that sat on a player-positive pack and test against the base rate.
Power — what a given sample can see
| Bets | Flag at | False pos. | Perfect | 75% | 50% |
|---|
Detector B — bet-log shape (no pack read)
Every window is late, so composition play is late. Share of a player's pair bets that fall in the last third of the shoe.
Watch a synthetic player get caught
Deal real seeded shoes, generate a player's pair-box log, run both detectors on it. Same shoes, only the player's policy changes.
pick a player, then deal.
Tell matrix (Appendix B, Lucky 6 direction corrected)
| Threat | Typical HE | Favourable state | Floor tell | Counter |
|---|
Detector C — the side-choice tell (known-card attack, §7.1)
The largest edge in the brief is not on a side bet: it is one known card in the next six, worth 6.4–9.1% on the main bet. It has the cheapest detector here, because the peeker must act on what he knows and the act is a side choice. A bettor with a live read bets the side the first card favours; a bettor without one cannot beat the ceiling.
| First card value | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|---|
| Favoured main bet | … | |||||||||
Power (α = 1%, exact binomial)
| Coups | Flag at | Perfect read | 3rd-card seat |
|---|
Cover: what it costs him to hide
| Misdirected | Match rate | Residual edge | Seen? |
|---|