The 14-Day and 30-Day Cliffs Are Releases, Not Cancellations

The lead-time histogram is the most boring chart on this site — until two bars refuse to behave. Openings detected on the day of the stay, tomorrow, and two to four days out all sit in the same neighbourhood (roughly four to five hundred each across the window), and the curve then decays smoothly through the second week: 150 at eight days, 127 at nine, 125 at ten, 127 at eleven, 125 at twelve, 99 at thirteen. Then, at fourteen days, 626. The next day, 52. The day after, 50. A process that produces 99 openings at thirteen days and 52 at fifteen does not produce 626 at fourteen by accident, and it does not produce it by cancellation either: cancellations have no reason to remember a fortnight. The thirty-day spike tells the same story at a second scale — 57 at twenty-nine days, 193 at thirty, 22 at thirty-one — a cliff, not a curve.

The fingerprints

A spike alone is a curiosity; the cohort's fingerprints make it a diagnosis. Fixed lead, daily cadence. The fourteen-day openings arrived day after day across the window, each day's batch sitting exactly fourteen days ahead of its stay dates — stay dates 2026-11-13 – 2026-11-17 in the current slice — the way a window that opens a fixed number of days ahead must. Uniform length. 586 of the 626 fourteen-day openings (93.6%) are single nights, against a 56.2% single-night share for the flow as a whole: the cohort is not a sample of cancellations, it is a shape. Concentration. Five facilities account for 569 of the 626 (90.9%): Rio Grande Village (Big Bend, TX) (153); Big Meadows Campground - Shenandoah (VA) (142); East Dunes Campground - Sand Lake Recreation Area (140); Cottonwood (Big Bend, TX) (94); Chisos Basin Campground (Big Bend, TX) (40). The rest of the system barely contributes to the spike at all — Rio Grande Village alone contributes 153 of its 280 openings at exactly fourteen days. The thirty-day cohort concentrates the same way: Kipahulu Campground (100); Crystal Lake (33); Hosmer Grove Campground (25) lead it, and Haleakala's two campgrounds together hold 125 of 193 (64.8%). Three fingerprints, all pointing one way: inventory entering the bookable pool on a schedule, facility by facility, window by window.

The analogy on this site is Cape Point, and the comparison is what makes the treatment defensible. Cape Point's pattern — near-identical daily batches, every alert at the same lead, essentially all distinct sites — was read as a rolling release in wave 3 and excluded from every cancellation count since, because its entire flow is the release: 996 alerts, 994 of them at exactly four days. The fourteen- and thirty-day cliffs are the same signature at a smaller share of their facilities' flow. Big Bend's campgrounds, Shenandoah's Big Meadows, East Dunes, Haleakala's pair — each of these also produces irregular, cancellation-shaped openings at other leads, in other lengths, on other days. Deleting their release cohorts from the counts would require a rule this site does not yet have a principled version of (how many days of fixed lead, at what concentration, before a cohort leaves the count?), and inventing one mid-series would break the comparability of every post that came before. So the treatment is disclosure: the cohorts stay in the counts, and this post — updated at every build from the same logs — is where they are named, sized, and subtracted by any reader who wants a release-free flow. That is a worse headline than a clean exclusion and a more honest dataset than a quiet one.

What a hunter does with a release cliff

The practical consequence is the opposite of the cancellation hunter's instinct. A cancellation is won by being fastest after it appears; a release is won by being present when it is scheduled to appear. If your target is Rio Grande Village, Big Meadows, East Dunes, Cottonwood, or Chisos Basin, the fourteen-day mark is a calendar appointment: the inventory for a given stay date enters the pool fourteen days ahead, in single nights, and the watch that matters is the one standing on that morning — the same morning-ritual logic as the 10 a.m. rolling-window routine, at a different hour and a different lead. If your target is Haleakala's Kipahulu or Hosmer Grove, the appointment is thirty days out. And the survival evidence does not punish patience here: fourteen-day-lead openings that completed a life cycle survived a median 12h 50m (n=223), against 4h 43m overall — a release cohort is not a flicker cohort, and its openings live long enough to be booked by a hunter who shows up on schedule rather than by reflex. The distinction to keep is the one this post exists to draw: do not read a fourteen-day spike as a surge in cancellations, and do not read a release as a reason to stop watching — read it as a second kind of opening, with its own clock, sitting inside the first kind's counts.

The limits, plainly

The release reading is an interpretation of a pattern, and this post labels it as such: the logs prove the fixed lead, the single-night shape, and the facility concentration; they do not contain the booking systems' release calendars, and no facility's calendar was consulted to confirm a fourteen- or thirty-day window. If a facility's window changes, the spike will move and this post's numbers will move with it at the next build — that is the test of the reading, and it is public. The window is six days (Oct 29, 10:54 PM ET – Nov 04, 05:33 AM ET) over November–December 2026 stays; other windows, other facilities, and peak season may show other cliffs, and the absence of a spike elsewhere in this window is not evidence that no other release windows exist — only that none produced a spike this large here. Counts are confirmed openings (two-stage verified). And the standing rule from wave 3 is unchanged: Cape Point's release is excluded from every cancellation count on this site, because its flow is the release; the fourteen- and thirty-day cohorts are disclosed, sized, and left in the counts, because their facilities' remaining flow is not. The live counts, cliffs included, are on the cancellation index; the counting method is on the methodology page.

Frequently asked questions

Last updated 2026-11-04.