The ruler now reacts to a volatility shock immediately (fast-up, slow-down)
and the rung is locked when the leg is born, so “To flip” can only fall
as price retraces — it can never move away from you mid-leg. Card 03 decides which
head is authoritative and dims the other. Neither predicts direction.
Runway leftclosed form · ρ 0.364
—
—
Awaiting data
New extreme first—
Another ½σ hour—
Quartile band—
vs 1σ hour—
Not a model — a formula. For a driftless walk under a
trailing stop, three numbers you already have (To flip, Retrace, Rung) give the whole
distribution of remaining travel. Tested against LightGBM on 552,271 rows: the formula
matched it (ρ 0.364 vs 0.359), so geometry is the ceiling. Median and the 25–75
band are well calibrated; the far-low tail reads slightly optimistic.
Side tiltv7 event clock · experimental
—
—
Awaiting data
The only place direction has ever been measurable. Built on a
separate clock: this card streams raw trades and closes a bar every fixed
$ of notional — the 1-minute data every other card uses carries no directional
signal at all (AUC 0.50). Validated strictly forward, trained on past only, with a
shuffle tripwire. It is a tilt, never a call. The edge decays with elapsed
time, so when the sub-line shows the live horizon far above the trained one, the
percentage is stale and should be ignored.
Record ledger
—
time
price
dir
rung
age
f15
f60
rng60
1σh
shift
retr
toflip
sinceX
eff
runway
pNew
flip in
zone
15m out
60m out
runway
कण kaṇa v5.1 — flip probability is a CLOSED FORM: z = ToFlip / (σ₁√T), P = 2Φ(−z), one monotone correction fitted on development data only. Sealed: flip15 AUC 0.826 / TRIGGER 0.708 at 6.19× · flip60 AUC 0.708. No flip model, nothing to retrain. Range head retained (ρ 0.710).