Which past nights your rules compare with (advanced)

Why it matters

A booking speed rule compares a night's recent bookings with what similar nights had received at the same point before arrival. The comparison is only as good as the nights behind it. This page is about how those nights are chosen, in the order MAYA tries things. Read Booking speed for what is done with them.

Which past nights count as similar does not depend on the rule's settings. What each rule counts on those nights depends on the room types it measures.

You never see the similar nights while a rule is being checked. You see them afterwards, under a price change in the change log or in Why this price?: open Show the numbers under a change on that night, then Which nights were compared?, then List the nights. Show the numbers is there only when at least one of your booking speed or booked so far vs usual rules was read on that night in the same run. Where the button appears has more.

Up to eight nights from three years

For a night that is not near a holiday, MAYA looks for at least 4 similar nights and keeps at most 8. They come from the three years before the night being priced, up to yesterday. A night near a holiday works differently: it is compared with that holiday in each of the three previous years, so it has 3 similar nights at most (read Holidays below). Some nights are never used, whatever else is true of them:

• nights inside a closed period you confirmed during the review
• nights you have set aside as not a fair comparison
• a window of dates around a yearly event, once your reports from 2 different years agree on it

Read Closed periods and set-aside dates below.

When too few nights match

For a night that is not near a holiday, MAYA looks back through your history, most recent first, in up to three steps. It takes the next step only when the ones before have found fewer than 4 nights.

1. First, the same weekday in the same season. MAYA steps back 7 days at a time and keeps the nights inside the same season, never a night near a holiday. The reasons you read are "a Friday", "your Peak Season" (or "your Year-Round"), and "earlier this year", "recently", "about a year earlier", "2 years earlier" or "3 years earlier".
2. If fewer than 4 are found, the same kind of night in the same season. Friday and Saturday are weekend nights, and every other night is a weekday night, so a Saturday can stand in for a Friday. The reason reads "a Saturday, similar to a Friday".
3. If there are still fewer than 4, the same weekday at a nearby time of year. Any night on the same weekday within 45 days of the same point in the year counts, whatever the season. The reason reads "a nearby time of year".

When your history is too thin to show seasons, or shows none, the whole year is one season called "Year-Round". The first step then keeps the nearest same-weekday nights, whatever the time of year.

When more than 8 qualify, the best 8 are kept: nights from the first step before the second, the second before the third, and within a step the nearest in time first.

If the second or third step was needed, the list under Which nights were compared? says what was added. After the second step the weekday line names the nights: "Too few other Friday nights matched, so Saturday nights were compared too, since they are also weekend nights. Friday and Saturday nights count as weekend nights, the rest as weekday nights." After the third, the season line says nights within 45 days of the same time of year were added. Either way a line follows: "Fewer than 4 nights matched closely, so the search was widened. Check the nights below and set aside any that were not normal." To check, open List the nights and look at the dates. If one of them was not a normal night, set it aside with Not a fair comparison? (read Setting a date aside).

Every night is read over the same stretch of its own booking curve as the night being priced. A Saturday 40 days before arrival is compared with how each similar Saturday was booking when it, too, was 40 days before arrival.

Weekends

Weekends are Friday and Saturday nights. Sunday is a weekday night. MAYA uses that in three places: in the second step of the search, where a Saturday can stand in for a Friday; in momentum, where a weekend night is read only from the weekend nights near it and a weekday night only from the weekday nights (see Momentum below); and in how seasons are read, which watches how Friday and Saturday nights book against the rest of the week (see Seasons below). Holidays use a wider weekend of their own: a holiday that falls on a Friday, Saturday or Sunday counts as a weekend holiday (see Holidays below).

"Our weekend is Thursday and Friday nights. Can MAYA use a different weekend?" No. There is no setting for which nights are your weekend. Away from holidays, the first and third steps of the search always compare a night with the same weekday, so your Thursdays are compared with Thursdays and your Saturdays with Saturdays. Only the second step, used when the first finds fewer than 4 nights, and momentum group nights into weekend (Friday and Saturday) and weekday (Sunday to Thursday). There a Thursday can be compared with a Monday, and a Friday with a Saturday. Every date on this page is a date on your property's own calendar. Read Time, weekends, currency and holidays.

Holidays

A night inside a holiday's stretch is compared by its position around the holiday, not by its weekday. The Saturday after Thanksgiving is compared with the Saturday after Thanksgiving in each of the last three years, whatever the dates were. Thanksgiving always falls on a Thursday, so there the weekdays line up.

For a holiday on a fixed date, the position around it is the same date every year, so the compared nights usually fall on other weekdays. Saturday, December 26, 2026 (the day after Christmas) is compared with Friday, December 26, 2025, Thursday, December 26, 2024 and Tuesday, December 26, 2023. All three are kept. The fixed-date holidays are New Year's Day, Valentine's Day, St. Patrick's Day, Juneteenth, Independence Day, Halloween, Veterans Day and Christmas.

A year in which the holiday fell on a similar part of the week scores higher, and the reason reads "similar weekend placement". The parts of the week are a weekend (a holiday on a Friday, Saturday or Sunday), a Monday or Thursday, and midweek (a Tuesday or Wednesday). The score only changes the order of the nights, never which ones are used. When the matching night sits nearer a different holiday that year, it is kept but ranked lower, with the reason "also Father's Day that year".

A night near a holiday has 3 similar nights at most, one from each year, and fewer when your history is shorter. With fewer than five, the reading is always kept within one step of Normal (read Where the lines fall). So on holiday nights booking speed reads Faster Than Normal, Normal or Slower Than Normal, never anything further out, and a rule set to a stronger level stays quiet there (Why a rule stays quiet). For holiday nights, use a rule set to Faster Than Normal or Slower Than Normal.

Under Which nights were compared? a line reads "This night is near Thanksgiving, so it is compared with the same days around Thanksgiving in earlier years, not the same date." Above it, the weekday line names the days of the week the compared nights fall on. Around a holiday that moves with the weekday, such as Thanksgiving, they are all the night's own weekday: "Only other Saturday nights were compared, since each night of the week books in its own way." Around a fixed-date holiday they usually differ, and the line lists them: "Around Christmas the day of the week changes from year to year, so the nights compared fall on Tuesday, Thursday and Friday." Under List the nights, each night's reasons give the year and the position, for example "Thanksgiving 2025; 2 days after the holiday; similar weekend placement".

MAYA has 17 holidays, all from the United States, and uses them for every property, wherever it is. Each has a stretch of days before and after it:

A night inside two stretches belongs to the nearer holiday, so December 31 and January 1 count as New Year's Day nights, and Christmas's own stretch runs December 24 to 30.

Nights that are not near a holiday are never compared with nights that are. A night near a holiday is compared only with the same holiday in earlier years, with one exception: if none of those past holiday nights has any bookings (for example in your first year), the night is read from momentum, which uses the ordinary nights around it. MAYA does not read weather, event listings or any holiday feed. Its only inputs are your own bookings.

"My hotel is outside the US. Do Thanksgiving and July 4 still change how my nights are compared? Can I turn them off?" Yes, they apply to every property, and there is no switch to turn them off. In a year, roughly 64 to 70 nights fall inside one of these stretches. Each of those nights is compared with the same position around the same US holiday in earlier years, and your ordinary nights are never compared with them. When none of those past holiday nights has any bookings, for example in your first year, the holiday night is read from momentum instead, which uses the ordinary nights around it. For a holiday on a fixed date, such as Independence Day (July 4), that means the same dates in earlier years, whatever weekday they fell on. For one that moves, such as Thanksgiving (the fourth Thursday of November), it means the matching days around it. On all of those nights, booking speed stays within one step of Normal, as described above. Setting those past nights aside takes them out of comparisons, but it does not make them ordinary nights. Read Handle a holiday MAYA does not have, which also lists the holidays outside the US that already fall inside a US stretch.

Seasons

"Season" here means a stretch of your year that books differently from the stretches around it: busier or quieter, with stronger or weaker weekends, or filling earlier or later. Your seasons come straight from your own booking history. They are worked out again each time your rules run, so you never set or maintain them (and cannot edit them).

Booked so far vs usual reads them from the 35 whole months before this one, so they stay the same all month and a night's similar nights don't change with nothing booked. A date you set aside still leaves them from the next run. Read Booked so far vs usual.

How seasons are read, in plain terms:

• The input is how many rooms each past night in the last three years sold, and how early it filled. Price is never used, so your own past pricing cannot bend the result.
• MAYA needs booked past nights on at least 120 different days of the year (about four months). With less, the whole year is one season called "Year-Round".
• A new season begins where the level, the weekly shape (how Friday and Saturday nights compare with the rest) or how early nights fill shifts and stays shifted for at least 10 days. A season is at least 10 days long.
• A blip of 7 days or fewer is smoothed away, however many years show it. A longer odd stretch survives only when at least 2 different years show it.
• Closed periods and every date you have set aside are removed first.

Seasons are named by their level: "Peak Season", "Low Season" and "Shoulder Season" (or "June Shoulder Season" when there are several). When the levels are close and the seasons differ in shape rather than size, they are named by their place in the year instead: "June Season" or "June to August Season".

You will only ever meet a season name inside an explanation, in a line such as "Only nights in the same season of your history were compared: Peak Season (June 10 through August 31)." or "Only nights in the same season of your history were compared: Year-Round (January 1 through December 31)." There is no seasons page and no season setting. Each change log entry keeps the season it used at the time. So an older entry can name a season, or season dates, that no longer match today's. New readings work seasons out again from your current history each time your rules run, and for booked so far vs usual on the first of each month.

Momentum, when nothing matches

Sometimes no similar night can be found at all: a room type that has never sold at this time of year, a holiday night whose earlier years have no bookings, a night far outside your history, or a night crowded by set-aside dates. MAYA does not invent a number. It makes one simpler reading, called momentum: how the nights near this one are filling right now, compared with how the same nights were filling a year ago. Only nights of the same kind count: a Friday or Saturday night is read from the Friday and Saturday nights near it, any other night from the Sunday to Thursday nights near it, because weekends and weekdays fill on different schedules.

In short: if the nearby nights have 6 bookings between them now, against 5 at the same point a year ago, they are filling about 20% faster. If this night itself had 4 bookings at this point a year ago, expected is 4 raised by 20%, about 5.

A momentum reading always counts as one similar night, so the level is capped at Faster or Slower. Read Booking speed for the cap.

Under Which nights were compared? the list opens with "Your history has no usable similar nights for this one, so the usual number comes from booking momentum instead." It goes on with lines such as "3 nearby nights could be paired with the same nights last year." and "Those nights are filling about 20% faster than they were a year ago.", and ends with "Momentum is rougher than comparing similar nights, so treat it as a best effort, not a measurement." The year-ago nights behind a momentum read are listed under List the nights, and you can set any of them aside like any other comparison.

Not enough history

Booked so far vs usual is stricter: it needs at least 3 similar nights with bookings and never uses momentum, so with fewer it isn't true. On a night near a holiday that means all three earlier years. Read Booked so far vs usual.

For booking speed, when there are no similar nights and momentum has fewer than 2 usable neighbouring nights of the same kind, the reading is "not enough history". A booking speed condition is never true on such a night, so booking speed rules do nothing there. Other standard rules (rules with no booking speed or pickup count condition, and no booked so far vs usual) are unaffected.

A rule that stays quiet leaves nothing in the change log, because the change log lists price changes. So you only see this reading when something else changed that night's price (another rule, a price you typed, or on Cloudbeds and ThinkReservations a rate change in your property system) while a booking speed rule was switched on. Then Show the numbers reads "There isn't enough booking history yet to say what is usual for this night." and "Booking speed wasn't rated for this night, so rules that watch booking speed left it alone."

The Rate Simulator shows it too: its Booking speed dropdown starts on Not enough history, so that booking speed rules stay quiet in a test unless you pick a level.

This happens on a night whose nearby nights of the same kind are nearly all holiday nights, set-aside dates or already past. A Friday or Saturday has fewer of them to start with, so it reaches this sooner, for example a Saturday near a holiday weekend. It also happens on a property with no bookings in MAYA at all, where no reading is made anywhere, so nothing is shown at all, not even under Show the numbers. A night far outside your history, or a room type that started selling recently, still gets a reading: when none of its similar nights has bookings, it is read from booking momentum, capped at Faster or Slower. On a new property it clears on its own once the first bookings arrive.

Booking speed rules also stay quiet when none of the room types a rule measures still counts as a room. Read Measure one set of room types, change another.

Closed periods and set-aside dates

Two kinds of night are never used as a comparison:

• Closed periods. Stretches of at least 14 days with no bookings (a stray booking or two allowed), with the property open on both sides, that you confirmed as a closure on the review after the import ("Were you closed ... ?"). Read Checking what MAYA found.
• Dates you set aside. Any night you flagged with Not a fair comparison?. It is left out from the next run. Say it happens every year and, once reports from 2 different years agree, a window of 5 days either side is skipped every year. Read Setting a date aside.

Undoing a set-aside date makes it comparable again on the next run.