2026-07-29 to 2026-08-28 · 26 sessions
Buying the mirror contract after a washout is a coin flip
After a same-day contract prints a violent low and recovers, buy the opposite side at the same strike. Sixty-five paper trades later it is break-even, and it does not survive dropping its best day.
In simple words
What happened
Across 65 positions over 26 sessions: 53.8% finished green, average +1.2%, median +3.0%. Resampling by day gives a 95% range of -7.1% to +9.2% — a range comfortably containing zero. Just over half the positions simply ran to the closing bell rather than reaching either the target or the stop.
Who would pay?
A violent flush that immediately recovers is often an engineered stop-run rather than a real move. If so, the side that just got washed out is the wrong side, and the mirror contract should pay.
How it was tested
Each qualifying washout produced one paper position in the opposite contract at the same strike and expiry, entered on a real two-sided quote and managed by a fixed rule: take profit at +50%, stop at -40%, flat by the close. Because the book is deliberately paper-clamped, every position ran to a genuine exit rather than being abandoned, which is why all 65 are gradeable.
What would prove it wrong?
Dropping the single best session takes the average from +1.2% to -0.8%. A result that changes sign when one day is removed is a description of that day.
Leader-collapsed evidence
What each setup actually did
65 paper positions, every one carried to a real managed exit — no unresolved rows
| Cut | Positions | Win rate | Mean | Median | Robustness | Decision |
|---|---|---|---|---|---|---|
| ALLAll mirror-contract entries | 65 over 26 sessions | +1.22% | +3.0% | 53.8% | 95% range -7.1% to +9.2% | REFUSED |
| DROP1Same set, best session removed | 62 | -0.81% | — | — | Sign flips on one day | REFUSED |
What we keep
Learnings
- 01
A win rate above half can sit on top of no edge at all. 53.8% sounds like something; with a +50% target against a -40% stop, average return is the number that decides, and it is +1.2%.
- 02
Drop-one-session robustness is cheap and it is the test that killed this. Any result that a single day can flip should be reported as a single day.
- 03
The paper clamp is what made the study possible. Because nothing was ever abandoned, all 65 positions have a real exit — the same lane graded on unresolved rows would have quietly mixed in defaults.
- 04
A lane can be worth keeping as a measurement while being worth nothing as a trade. The cost of running this one is the alert surface it occupies, and that is a real cost.