A nine-bar succession count, and the one part of it that survived a walk-forward test.
A succession count marks each bar that closes beyond a fixed reference a set number of bars back. Nine in a row is the classic exhaustion signature: a trend that has extended without a single pause is a trend running on fewer and fewer participants.
The validation level (the close of the last zero-succession day) was tested walk-forward over 480 events. It was reached 44.2% of the time against a 30.1% base rate: a 14.0 point edge, with a 95% confidence interval running from +8.9 to +19.5 points. That interval clears the base rate, which is what separates a result from a coincidence.
Most of them. The 7.86% and 10% primary targets do not separate at any window we tested. The deep -16.18% retracement was reached once in 477 events. The arrival-window effect at days 16 to 19 has a confidence interval spanning zero, which means it is not there. We publish all of it, because a product that only shows its wins is a brochure.
The count anchors on the highest high of day four or five after the setup, not on the swing high a chart package would draw. Those differ more often than you would expect, and the measurement above is only valid for the definition it was run against.
A cycle validates when price closes below the validation level inside the day 8 to 9 window. When it does not — one recent cycle missed by 0.77 cents at the close — the selloff thesis failed at a level the market defended, and the framework reads that failure as a long: buy the level, stop under the low the cycle made getting there, target the swing high the ladder was measured down from.
Two findings, and the first is about availability rather than accuracy. Of 276 cycles that failed to validate across SPY, QQQ, IWM and DIA over ten years, 235 had already closed above the swing high by the time the window shut — there is no setup, because the target is behind price. In the 41 where the target was still ahead, the setup reached it before its stop 48.8% of the time against a matched null of 52.8%. That is a lift of minus 4.0 points on an interval running from minus 17.3 to plus 8.2, and eight different parameterisations all came back negative. It is published because it was built, not because it works.
A cycle targets its swing high, which is a price the market has already printed. So the target can never sit above the record high — at most it can BE the record high, and only in that case is reaching it the same event as printing a new one. Among the setups that did reach target, about one in ten saw a record high before the position closed. The surface states which of the two it means, every time.
A bar count that marks a trend running out of participants. Each bar closing beyond a fixed reference a set number of bars back advances the count, and nine consecutive is the classic exhaustion signature — a trend that has extended without a single pause.
One part of it does, on the measurement we ran. The validation level was reached 44.2% of the time against a 30.1% base rate over 480 walk-forward events, a 14.0 point edge whose 95% confidence interval runs from +8.9 to +19.5 and therefore clears the base rate. Most of the rest did not survive: the 7.86% and 10% primary targets do not separate at any window we tested, and the deep retracement was reached once in 477 events.
Because a product that only shows its wins is a brochure. Every measurement is at /proof with its sample size, its interval and the detector fingerprint it was produced under, including a measurement of a figure a competitor publishes that came back flat.
Validation requires a close below the validation level inside the day 8 to 9 window. Without one the cycle is invalidated: the selloff it was counting did not arrive. The framework reads that as a long setup entered at the defended level, stopped under the cycle low and targeting the swing high. We measured that setup and it does not beat a matched null.
Not on our measurement. Across 41 cases where the target was still ahead of price, it reached target before stop 48.8% of the time against a ticker- and month-matched null of 52.8% — a lift of minus 4.0 points whose 95% interval runs from minus 17.3 to plus 8.2 and therefore contains zero. None of eight parameterisations of horizon and tie-break came back positive.
Because 235 of them leave no setup at all. An invalidated cycle is one where the selloff never happened, so by the time the validation window closes price has usually already gone back above the swing high the cycle was measured down from. The target is behind price and there is nothing to recover to. Only the remaining 41 were takeable, and those are what the rate is computed on.
Usually not. The target is a prior swing high, so it can only equal the record at best, never exceed it. Of the setups that reached their target, roughly one in ten printed a record high before the position closed. When a cycle’s target happens to BE the highest high in the data, reaching it and setting a record are the same event — and that is the only case where the two can honestly be quoted together.
Dealer gamma across the expiry curve, and a written read every session.
CREATE A FREE ACCOUNT, NO CARD