Cite as: SnagRes Reservation Availability Dataset, snagres.com/data, as of 19 October 2026. Open dataset. Release schedule.
Two places, and they are different. Release times and booking windows come from each restaurant's own published rule, read from its booking listing or website, or for a few, a third-party write-up, which the release schedule marks as reported. Where a restaurant publishes no time, the release schedule shows the hour we have repeatedly seen new dates appear and labels it as observed. Availability percentages come from repeated checks of each public booking calendar, recording which dates and times were bookable for parties of two and four.
The share of the visible booking window that had at least one bookable slot at the last check. A restaurant showing 10% had tables on roughly one day in ten across the window, which is why a low figure means an alert is worth more than refreshing.
A time labelled as the restaurant's rule is taken from the restaurant's own statement or its platform listing. A time labelled as observed is the hour new dates appeared across several days of checks, and is published only once it repeats. Where a restaurant states how far ahead it books but not at what time, we say the hour is not published rather than guess one.
Cite it as the SnagRes Reservation Availability Dataset (snagres.com/data) with the date you read it, since the figures are refreshed as new checks come in. The data is licensed CC BY 4.0, so any use is fine with attribution.
Figures measured 19 October 2026. Recheck cadence varies by city, platform and demand.
Every restaurant page on SnagRes carries statistics about how that specific restaurant books. This page explains where those come from, so you can judge how much weight to put on them.
SnagRes checks the booking calendar of restaurants in New York City, San Francisco, Los Angeles and Chicago across Resy, OpenTable, Tock and SevenRooms. For each restaurant we record every date in the visible booking window, whether a table was available on that date, and the exact time slots offered, separately for a party of two and a party of four.
That produces a continuous record rather than a sample. The figures currently on the site rest on 61,530,416 individual date checks.
These are the times left open, not the restaurant's service hours. At a hard restaurant those are close to opposites, because the desirable times never survive. The slot that opens most often at Carbone is 11:30 AM, and only 0% of its logged openings fall between 7:00 and 8:30 PM. "The slot that opens most often is 11:30 AM" is true. "Carbone usually seats at 11:30 AM" is false, and we are careful never to write the second.
"Never observed" is a statement about our observations. When we say we have never seen a Sunday table at a restaurant, that is what we measured. It normally means the restaurant is closed on Sundays, and we say so, but it is not a statement about their published schedule.
Every figure on this site, one row per restaurant. Free to use for any purpose under CC BY 4.0: credit SnagRes and the date you read it. If you are checking our work, start here rather than with the summary numbers above.
Download the full dataset (CSV) ↓
“Open” is the share of dates in the visible booking window that had a table for two at the last check. It is a reading of the calendar as it stands, not a history of the last month, and the two answer different questions.
If a figure here is wrong about your restaurant, write to [email protected] and we will check the underlying record and correct it. SnagRes reports availability it observed; it does not rate restaurants, and it never books on anyone's behalf.
Every booking statistic on this site is something we recorded, not something we estimated. This page shows the working: what we check, how often, what each figure means, and the two ways it is easy to misread.
Of the dates visible in the 30-day booking window at the last check, the percentage that had at least one table available. Reported separately for parties of two and four, because at many restaurants they differ substantially.
The share of all observed open slots that fell between 7:00 and 8:30 PM. A rate of 0% means that across every slot we have ever recorded at that restaurant, none was in that window.
The single most frequent start time among slots that were open. See the warning below: this is not the restaurant’s most common seating time.
The weekday with the highest and lowest share of checks that found availability. Days on which we have never seen a table are excluded from this comparison and reported separately, because a closed day is not a hard day.
A weekday with a substantial number of checks and no recorded availability at all. This usually indicates a closing day.
We see the public booking calendar. Tables held back for walk-ins, the bar, the phone or a concierge are invisible to us, so a restaurant showing 0% may still seat you another way.
The window is 30 days. A restaurant that books 90 days out has most of its calendar outside our view.
A figure is only published once enough observations sit behind it. Restaurants below that floor get a shorter page rather than a shakier number, which is why some pages carry fewer statistics than others.
Coverage is uneven. The busiest rooms carry several times the observations of the quietest, and the table above shows how the sample splits across platforms so you can judge whether any single one is carrying a finding.
Release-frequency figures are reported alongside the window they cover. Our change detector was corrected on 17 August 2026 and everything recorded before that is discarded, so that window is currently short and the pages say so rather than implying a month of history.
Keep reading: New York tracker · What SnagRes is