🍝 When restaurant tables actually open up
The advice for booking an impossible restaurant is always the same: set a reminder for the day the calendar drops, be fast, try again in a month. It treats the reservation as a race you enter once.
That is not how the inventory behaves. We have been watching, and tables come back constantly — they just come back on a schedule nobody publishes.
Since 13 July 2026 our scouts have checked 21 hard-to-book restaurants roughly every 15 minutes. Through 8 August they caught 353 tables reopening. Here is what the pattern looks like.
Openings cluster in the 8am hour
Of the 353 openings, 82 — 23% — appeared during the 8 o'clock hour, local to the restaurant. No other hour comes close: the runner-up is 2pm with 26.
That is a threefold gap between first and second place, and it holds across cities. If you are refreshing by hand, refreshing at 8am is worth more than refreshing all afternoon.
There is a second, much softer pattern late at night — midnight and 10pm both show around two dozen openings. That is the shape you would expect from people cancelling before bed rather than from anything systematic.
Half of them are gone in three and a half hours
For 124 of those openings we saw both the moment the table appeared and the moment it disappeared. The median gap was 3 hours 28 minutes.
The distribution matters more than the median:
| Gone within | Share of openings |
|---|---|
| 15 minutes (one check) | 14% |
| 30 minutes | 27% |
| 2 hours | 40% |
So roughly a quarter of returned tables are gone within half an hour, but a meaningful tail sits open for most of a day — the 75th percentile is just under 16 hours. A table that survives its first hour will often survive until evening.
This is the part that makes manual checking so unrewarding. The openings are frequent, but the ones worth having are frequently short, and they do not wait for you to check again after lunch.
Cancellations land about eleven days out, not six months
When a table reappeared, we recorded which night it was for. The median gap between an opening appearing and the night itself was 11 days. A quarter of them were for a night within 3 days.
That is the finding most at odds with conventional advice. "Book six months ahead" describes the initial release. It does not describe the cancellation flow, which is where most of the returned inventory actually lives — and that flow is concentrated inside the next two weeks.
If you want a table next Thursday, next Thursday is a more realistic target than it sounds.
Thursday is the hardest night
Counting the nights that opened back up, weekday evenings dominate:
| Night | Openings |
|---|---|
| Thursday | 59 |
| Wednesday | 55 |
| Friday | 52 |
| Saturday | 52 |
| Tuesday | 50 |
| Sunday | 31 |
| Monday | 30 |
Read this carefully, because it is easy to read backwards. More openings on Thursday does not mean Thursday is easier — it means Thursday is where the volume and the churn both are. Sunday and Monday show half the movement, which is what a genuinely quiet night looks like from the inside.
Where the openings came from
Not every room behaves the same way. Across the four weeks:
| Restaurant | Openings caught |
|---|---|
| Carbone (Miami) | 59 |
| COTE (New York) | 38 |
| Lilia | 28 |
| Peter Luger | 22 |
| Zahav | 22 |
| Funke | 21 |
| Bridges | 17 |
| Sunny's | 17 |
We are not ranking these as "easiest." Sample sizes at the bottom of that list are small, rooms differ in size, and four weeks in high summer is one slice of one season. Treat it as where the movement was, not as a league table.
What we cannot tell you
Our cadence is about 15 minutes. Anything that appeared and vanished inside one interval is invisible to us, so every number here is a floor. When we say 14% of tables were gone within 15 minutes, the honest statement is "gone before we looked again" — the real figure is faster and we have no way to measure it.
Three of the restaurants we watch logged no openings at all in this window. That may mean nothing came back, or it may mean we are not seeing their inventory correctly. We would rather say so than quietly drop them from the denominator.
And the counts above come from the change log we can currently retrieve, which does not reach all the way back to when watching began. The totals are conservative.
Four weeks is not a season. We will publish this again with more behind it.
---
These findings come from Calafia scouts watching public reservation availability. Each restaurant we watch has its own page with its live check log — for example Via Carota — and you can have one email you the moment a table opens. We alert; the booking stays in your name.