How is the NFL schedule made? The formula, then the dates

Every spring a schedule is released, several million people open it at once, and within an hour somebody has published a list of which club has the hardest year ahead. Almost all of that analysis is aimed at the wrong object. By the time the schedule appears, the fixtures have already been settled for months. What was released is the order.

That distinction is the whole answer to how is the NFL schedule made, and nearly every explainer blurs it. There are two entirely separate processes with different owners, different timings and different degrees of discretion. The first decides who plays whom, and it is a formula with no judgement in it at all. The second decides when each of those games is played, and it is one of the hardest optimisation problems any sports league sets itself.

Get the two apart and everything else falls into place, including why the annual complaint about an unfair schedule is usually a complaint about arithmetic.

Your opponents were fixed before anybody sat down to schedule anything

When the final whistle of the regular season goes, the following season's 272 fixtures already exist. Not as a plan, not as a draft. As a determined set. The last game finishes, the standings are final, and the formula produces the complete list of who plays whom with nothing left to decide.

This is why the league is able to publish each club's opponents for the coming season within days of the previous one ending, months before the schedule itself. Those two announcements are constantly treated as the same event by supporters, and they are not remotely the same event. One is the output of a calculation. The other is the output of a search.

The formula divides a club's 17 games into five groups. Six against division rivals. Four against another division in the same conference. Four against a division in the other conference. Two against clubs elsewhere in your own conference, matched on last season's finishing position. One against a club in the other conference, matched the same way.

35%24%24%12%

• Division rivals, home and away6
• Rotating division, same conference4
• Rotating division, other conference4
• Same-conference games matched on last season's finish2
• The seventeenth game, other conference, matched on finish1

Set by the league's published scheduling formula. Every club in the league has exactly this composition, every season.

Show the numbers

Where a club's 17 games come from

Item	Value
Division rivals, home and away	6
Rotating division, same conference	4
Rotating division, other conference	4
Same-conference games matched on last season's finish	2
The seventeenth game, other conference, matched on finish	1

Six plus four plus four plus two plus one is seventeen, which is the entire card. There is no discretionary game, no exhibition slot, no fixture held back for a marquee pairing somebody wants to sell. If a matchup you were hoping for is not on the list, no amount of television appetite will put it there.

Count the distinct opponents rather than the games and you get fourteen, because the three division rivals appear twice each. So a club faces fourteen different opponents out of the other thirty-one, and which fourteen was determined by two rotations and by where everybody finished.

The two rotations do the heavy lifting, and they are why the fixtures feel fresh

Eight of the seventeen games come from block rotations, and the design of those rotations is the most elegant part of the whole system.

Each conference has four divisions of four clubs. In any given season, one division plays a different division in its own conference in full, all four clubs against all four clubs, two at home and two away. There are three other divisions in your conference, so that rotation comes round on a three-year cycle. Every third season you play the same set again.

The interconference rotation works identically. Your division is paired with one division from the other conference and plays it in full, two home and two away. There are four divisions over there, so that cycle is four years long.

The consequence is worth stating plainly, because it is the thing the rotation was built to guarantee. Within four seasons every club plays every other club in the league at least once. Nobody can go a decade without a fixture, and no rivalry can be quietly avoided.

Play another division in your own conference3

Play a division in the other conference4

Face every club in the league at least once4

Guaranteed to play a game outside the United States8

Structural properties of the published rotation. The international guarantee is a league commitment rather than a rotation, and is included because it is the longest cycle a club sits inside.

Show the numbers

How long each cycle takes to come round, in years

Item	Years
Play another division in your own conference	3
Play a division in the other conference	4
Face every club in the league at least once	4
Guaranteed to play a game outside the United States	8

There is a hidden fairness property in the block design that rarely gets credit. Because a whole division plays a whole division, all four clubs in your division get precisely the same eight opponents from the rotations. Not comparable opponents. The same ones. If your rivals' schedules look easier, it is not the rotation doing it.

Three games, and only three, are decided by how good you were

The remaining three fixtures are the league's one concession to competitive balance, and they are far smaller than people assume.

Two of them are within your conference. Take the two divisions in your conference you are not playing in full this season, find the club in each that finished in the same position as yours, and there are your opponents. Division winners play division winners. Fourth-placed clubs play fourth-placed clubs. One of those games is at home and one is away.

The third is the seventeenth game, and it applies the same idea across the conference line. A club from a division in the other conference that is not already on your card, matched to your finishing position.

So a club that wins its division is handed three extra games against clubs that also won theirs, and a club that finishes bottom is handed three against clubs that also finished bottom. That is the entirety of the handicap. Three fixtures out of seventeen, or a shade under eighteen per cent of a season.

Compare that with how much of the schedule the formula hands you regardless of merit and the proportions become clear. Fourteen games are set by structures nobody chose this year. Three respond to how you played. A league that genuinely wanted to level results through the schedule would weight it far harder than this, and the reason this one does not is that the block rotations are doing a different and more valuable job: they make the standings comparable, because clubs in the same division are being measured against almost the same opposition.

The seventeenth game is an odd number, and odd numbers break home and away

Seventeen games cannot be split evenly into home and away. Somebody gets nine and somebody gets eight, and the league resolved that by moving the imbalance up a level.

The seventeenth fixture is always interconference. In a given season one conference hosts all of them and the other conference travels for all of them, and the following season it swaps. So a club has nine home games one year and eight the next, and over any two seasons the count comes out even. The league publishes which conference hosts in which season, and that is the document to check rather than any summary of it.

That arrangement has consequences that are rarely thought through. A club with nine home games is playing an extra date in front of its own crowd, which is worth something real in the standings and considerably more at the box office. A club chasing a division title in an away year is doing it with one fewer home fixture than it will have next season, and its rivals in the same division are in exactly the same position, which is the point. The imbalance is applied to whole conferences, never inside a division, so it can never distort a division race.

The interconference rotation completes on a four-year cycle, and the seventeenth game adds one more crossover fixture on top of it. The league has said the arrangement gets revisited once the cycle has run its course, and if the format changes it will change in the published formula first. Treat anything else you read as a proposal.

Deriving one club's card from the formula, step by step

Here is the calculation run end to end for an invented club. Nothing below describes a real season, and every name is made up.

1. 1Start from last season's final standingsArdwick Foundry finished second in its division. That single fact, plus the calendar year, determines every opponent below. No club, broadcaster or league official contributes anything at this stage.
2. 2Six games against the three division rivalsHome and away against each of them. These are fixed for ever and never rotate, which is why divisional fixtures are the only games guaranteed to appear on every schedule your club will ever play.
3. 3Four games against the rotating division in your own conferenceThe whole division plays the whole division, two at home and two away. All four Ardwick division clubs receive these same four opponents. Which division it is comes from a three-year cycle, not from anybody's preference.
4. 4Four games against the rotating division in the other conferenceThe same block arrangement across the conference line, two home and two away, on a four-year cycle. Again identical for all four clubs in Ardwick's division.
5. 5Two games matched on finishing position, same conferenceArdwick finished second, so it plays the second-placed club from each of the two conference divisions it is not facing in full. One at home, one away. Kestrel Bay is one of them purely because it also finished second.
6. 6One game matched on finishing position, other conferenceThe seventeenth fixture. A second-placed club from an other-conference division not already on the card. Saltmarsh Athletic qualifies on that basis alone, and the hosting conference for the season decides whether Ardwick travels.
7. 7Seventeen games, fourteen opponents, nothing left overThe formula has accounted for the whole card with no residue and no gaps. At this moment not one date, kickoff time or television window exists. The schedule has not been made. The fixtures have.

Ardwick Foundry, Kestrel Bay, Saltmarsh Athletic and the divisions named around them are invented. The sequence of steps is the league's published formula; the clubs and their finishing positions are not real.

Read that flow again and notice what is missing from it. No stadium. No network. No week number. No bye. The formula produces a set of unordered pairings, and every hard question in scheduling begins immediately afterwards.

The formula makes the schedule mostly balanced, and here is exactly where it fails

The balance argument is usually made loosely, so it is worth doing precisely.

Take two clubs in the same division. Their six divisional games are the same six fixtures viewed from opposite sides, since they play each other twice and both play the other two rivals twice. Their eight rotation games are against identical opponents. That is fourteen of seventeen games against the same opposition. Only the three finish-matched fixtures can differ, and those differ only because the two clubs finished in different positions.

Fourteen out of seventeen. When a supporter says a division rival got the softer schedule, the largest possible gap between them is three fixtures, and those three were earned rather than assigned.

Between clubs in different divisions, the picture is completely different, and this is where the imbalance genuinely lives. Two clubs in different divisions of the same conference share almost nothing structurally. Different divisional opponents, different rotation partners in most years, different finish-matched games. Their records are built against substantially different opposition, and no amount of formula design fixes that, because it cannot be fixed without making everyone play everyone, which needs a season more than twice as long.

That is exactly why the tiebreakers that decide postseason places lean so hard on head-to-head results and common opponents before they will look at raw records, and why the wild card sequence handles common games differently from the division sequence. The schedule formula creates a specific pattern of comparability, and the seeding rules are built on top of that pattern rather than pretending it does not exist.

There is a second failure mode worth naming, because the annual strength-of-schedule articles depend on it and never mention it. Opponent quality in May is measured using last season's records. Those records are a lagging indicator of clubs that have since drafted, traded, signed and in several cases replaced their coaching staff. A published list of the hardest schedules is a prediction wearing the clothes of a measurement, and its accuracy in September is roughly what you would expect from any list built on twelve-month-old data.

Ordering the fixtures into weeks is the part that is genuinely hard

Now the real work starts. Two hundred and seventy-two fixed matchups have to be placed into eighteen weeks, with each club playing seventeen of them and resting once, and every placement has to satisfy a stack of constraints that mostly have nothing to do with football.

• 272Fixtures to be placed
• 18Weeks to place them in
• 32Byes to be positioned
• 17Games each club plays

Structural counts. 32 clubs at 17 games each is 544 club appearances, which is 272 fixtures. 32 clubs across 18 weeks is 576 club weeks, leaving exactly 32 byes, one per club.

The constraints arrive from several directions at once, and they contradict each other.

Stadium availability comes first and is non-negotiable. The league collects conflicts from clubs early in the calendar year: concerts already sold, other tenants with priority, civic events, work on the playing surface. A stadium that is unavailable on a given weekend removes every schedule in which that club is at home that weekend, and the effect ripples, because the opponent now needs a different home date too.

International fixtures are placed before almost anything else. Games played outside the United States occupy fixed slots, a host club gives up one of its home dates, and the schedule makers then work a window of about three weeks around each one to soften the travel: which fixture precedes it, whether a bye follows it, how far each club has to come. Every club is guaranteed to play in one within an eight-year period, so the pool of candidates is constrained by history as well as by appetite.

Byes have to be distributed rather than allocated. One per club, placed inside the permitted window, and spread so that no week is left with too few games to fill the television slots that week requires. A bye in one week is a bye not available in another, for both the club and everybody it plays around that date.

Short weeks carry a competitive cost the league tries to contain. A Thursday fixture follows a Sunday one, so both clubs lose most of a normal preparation week. That is tolerable on its own and much less so stacked on top of travel. Two patterns in particular count heavily against a candidate schedule: a road game on Monday followed by another road game the next Sunday, and a Thursday road trip landing on the back of a long journey the previous weekend. These are not absolute prohibitions. They are penalties, and a schedule containing a few of them can still be the best one available.

Travel is managed in runs rather than in totals. The league tries to hold consecutive road games to two, with particular attention at the start and the end of the season, and sometimes cannot. Cross-country trips are spread out rather than clustered. Time zone crossings matter more than raw distance, which is why a short hop north is nothing like a flight of the same length westwards.

Broadcast windows have to be filled, and they have shapes. A fixed number of national windows exist each week, some of them exclusive, and every one of them needs a game worth putting in it. Clubs cannot all appear in prime time equally, some carry floors and ceilings on appearances, and a Sunday night window standing alone must hold an audience with a single fixture.

A handful of dates are effectively pre-booked. The opening Thursday, which the league usually gives to the reigning champion at home. The Thanksgiving games. The Christmas slate. Each of these takes a fixture out of general circulation and constrains the two clubs involved on either side of it.

Weather sits underneath all of it. Outdoor stadiums in northern cities are a different proposition in late December from what they are in September, and roofed venues can absorb late dates that others cannot. This is never stated as a rule and it shapes a great many placements.

The final week is reserved for divisional fixtures. Every club finishes against a rival with a direct stake in the same standings, which closes off the situation where a club plays a meaningless final game while a competitor plays a live one.

Why this is a combinatorial optimisation problem, not a puzzle somebody solves

Take the constraints away for a moment and just count. Two hundred and seventy-two fixtures, eighteen possible weeks for each. That is 18 to the power of 272, a number of 342 digits, and it is an upper bound rather than a count of valid schedules, because the overwhelming majority of those assignments are nonsense: a club playing twice in a week, a week with thirty games in it, a bye that never happens.

The point of the number is not its size. It is that the search space cannot be enumerated by any computer that will ever exist, so the schedule is not found by checking every option and picking the best. It is found by searching.

What stops the problem decomposing into thirty-two easy problems is that every fixture involves two clubs. Move a game to solve one club's travel run and you have altered the other club's week, which may break its bye placement, which may empty a television window, which may force a third fixture to move. The constraints are coupled across the whole league, which is the property that makes scheduling problems of this family computationally hard in the formal sense. Nobody proves the released schedule is optimal, because proving it is not achievable.

So the league does what is done with every problem of this shape. It encodes the constraints, generates enormous numbers of candidate schedules on cloud computing, scores each one against an objective that penalises the patterns it wants to avoid and rewards the ones it wants, and keeps the best of what it finds. The published description is thousands of computers producing thousands of viable schedules, which is a compression of a much longer process of successive refinement.

Then a small group of human beings picks one.

That last step is the part people find hardest to believe and it is the most defensible thing in the process. An objective function is a set of weights, and weights are opinions about how bad a Thursday road trip is compared with an empty November window. No weighting captures every consideration, and the final choice between a handful of good candidates involves judgements about which club has a genuine grievance and which merely has a preference. Machines narrow the field from unimaginable to a shortlist. People choose from the shortlist and take the criticism.

1. 1Begin with 17 fixtures and no datesArdwick's opponents are already final and cannot be changed by anything that follows. What is open is only which week each game falls in, which day, what kickoff time, and where the bye sits.
2. 2Lock the immovable dates firstThe opening Thursday, the international slots, Thanksgiving and Christmas. Ardwick is in none of them this year, but two of its opponents are, which removes several weeks from those fixtures before Ardwick's own preferences are considered at all.
3. 3Apply the stadium blackoutsArdwick's ground is unavailable for two autumn weekends because of events booked before the season existed. Every candidate schedule with Ardwick at home in those weeks is discarded, and its opponents in those fixtures need new home dates as a result.
4. 4Place the bye, then place it againA mid-season bye suits Ardwick. It also suits eleven other clubs, and the week can only carry so many. The bye moves, which shifts the two fixtures either side of it, which alters an opponent's travel run three weeks later.
5. 5Score the travel and short-week patternsOne candidate gives Ardwick a Monday road game followed by another road game the next Sunday. That is heavily penalised and the candidate is set aside, even though it beat its rivals on every television measure.
6. 6Fill the national windowsArdwick's fixture against Kestrel Bay is attractive and lands in a prime-time slot. That decision constrains both clubs on either side of it, because a night game changes the recovery pattern of the following week for everyone involved.
7. 7Choose between the survivorsSeveral thousand valid schedules remain, all defensible and none perfect. Humans compare the shortlist, weigh whose complaint is legitimate against whose is a preference, and pick one. Ardwick's opponents are exactly the ones the formula produced in January.

Ardwick Foundry and its opponents are invented, and so is every week placement below. The constraint types and the sequence of stages follow the league's published process; the specific fixtures do not describe a real season.

Flexible scheduling changes the window, never the fixture

Once the schedule is out, one more layer of change exists, and it is narrower than its reputation.

Flexible scheduling lets the league and its broadcast partners move an already scheduled game into a different slot. The Sunday night window can be flexed a limited number of times in the middle stretch of the season and more freely late on. Monday night and Thursday night both become eligible in the closing weeks. Sunday afternoon games can slide between the early and late windows without any of the ceremony, which happens constantly and is barely noticed. The final week is treated separately: the fixtures are known all along, because they are divisional games, but the days and kickoff times are set only once the previous week has been played and the standings make clear which games decide something.

What flex cannot do is the point. It cannot change who plays whom. It cannot change the venue. It cannot move a fixture into a different week. It moves a game between windows inside the week it was always going to be played in, and minimum notice periods apply so that supporters holding tickets and travel are not stranded. Those periods are published by the league and differ by window.

The reason flex exists at all is a forecasting problem nobody can solve. A prime-time slot is sold months in advance on a prediction of which fixtures will matter in December, and predictions about which clubs will be good are wrong at a rate that would embarrass anybody who had to publish a record. Before flex, a national window in week fifteen could be occupied by two clubs long since eliminated while a genuine race played out in front of a regional audience. Flex is the mechanism for correcting a forecast after the fact, and the money involved in selling those windows in the first place explains exactly why the league wanted it.

There is a cost, and clubs feel it. A club whose game is flexed loses the day it planned around, and so do its supporters. The compensation is that the club is in a game that matters enough to be moved, which is a form of consolation only available to clubs having a good season.

Fairness against television value, stated honestly

Here is the tension without the diplomacy.

The formula is objective and nobody argues about it, which is precisely because it contains no discretion at all. Once the standings are final, the fixtures follow, and no broadcaster, owner or league executive can add a matchup they want or remove one they do not. That is a genuinely strong design and it deserves saying clearly before anything critical follows.

The ordering is where money enters, and it enters heavily. National windows are the most valuable inventory the league sells, and the schedule makers are explicitly trying to fill them well. Clubs are not treated equally in that process. A club with a national following, a compelling quarterback or a recent run of success will be moved into prime time more often than a club without those things, and the league does not pretend otherwise.

The competitive cost is real and it is measurable in a way the benefit is not. A night game shortens the following week. A Thursday fixture shortens the week before it. Clubs that appear in prime time repeatedly absorb more of those disruptions than clubs that do not, which means the most watchable clubs are handed the most awkward calendars. The reward for being interesting is a worse schedule.

The league's defence is structural rather than moral, and it is stronger than it first sounds. National broadcast revenue is shared, so a club that gives up a comfortable Sunday afternoon for an inconvenient Thursday is contributing to a pot every club draws from, including the ones that never get moved. That pot is what sets the ceiling on what clubs may spend, so the same money that produces the awkward Thursday also produces the limit that keeps the richest clubs from simply buying the best roster. Take the television value out of the ordering and you take money out of a system deliberately built to redistribute it.

Where the defence gets thinner is in the concentration. Sharing the revenue does not share the fatigue. A club that plays four night games and a Thursday road trip has absorbed a competitive cost that a club with none of those has not, and it receives the same share of the money either way. The league manages this with caps and floors on appearances rather than with a proper accounting of the burden, and that is a policy choice rather than a solved problem.

It is worth noticing how different this is from competitions that hand fixture selection to a draw. In the format where clubs are drawn into a single league phase the opponents themselves come out of a randomised process that has to be constrained afterwards to stop implausible pairings. Football's approach randomises who plays whom and accepts what it gets. This one determines who plays whom and then optimises when, which concentrates all the discretion in one place, where it is easier to see and easier to criticise.

Why the complaint about a hard schedule is usually a complaint about the formula

Every spring produces the same argument, and almost all of it is misdirected.

If the grievance is about the quality of the opponents, it is not about the schedule at all. Fourteen of the seventeen came from the division and the two rotations, which were determined by a cycle and by an alignment decided decades ago. The remaining three came from finishing where you finished. A club that wins its division and then complains about facing three division winners is complaining about having won, and a club stuck in a division with three strong rivals is describing a structural fact that will hold for as long as the alignment does.

If the grievance is about the number of good clubs on the card, check who your division rivals are playing. If they are in your division, fourteen of their games are against the same opponents as yours. The comparison people actually want to make is with a club in another division, and that comparison is legitimate but it is a criticism of the rotation rather than of the schedule makers, and the alternative to the rotation is playing fewer opponents less often.

There is a narrow band of complaint that lands, and it is worth knowing what it looks like, because it is where the real discretion sits. Three road games in a row. Two long trips separated by a short week. A bye in week five followed by thirteen straight games. Facing three consecutive opponents who are all coming off byes while your own bye sits somewhere useless. Those are ordering decisions, they were made by people, and the league's own schedule makers have acknowledged in public that some clubs get a fair hearing on them. That is a much smaller complaint than the one usually made, and it is the only one with a person on the other end of it.

There is also a grievance that is invisible until you look for it. A club's opponents may all be faced at their strongest or all at their weakest, purely as a function of when in the season those fixtures fall. Play a club in week two and you may catch it before its injuries; play the same club in week sixteen and you may catch it resting people because its position is already secure. The formula cannot see this and the ordering cannot control it, and it probably moves more results than every travel constraint combined.

Reading a released schedule properly

When the schedule appears, most of what is in front of you is old news presented as new. Here is how to separate the two.

Sort the seventeen into the five buckets first. Six divisional games, four in one rotation, four in the other, two conference games matched on finish, one interconference game matched on finish. Every fixture belongs to exactly one bucket. If your total does not come to seventeen, you have miscounted a divisional game.

Identify the two rotation divisions and note that they are not news. They were determined by a cycle, and every club in your division has the same eight opponents from them. Any comparison you make with a division rival about opponent quality is a comparison about three games.

Look at the three finish-matched fixtures and check your own finishing position against them. If they look hard, your club finished high. If they look soft, it did not. This is the only part of the opponent list that responds to anything your club did, and it responds in the opposite direction to how supporters usually read it.

Now start reading the ordering, because that is the new information. Where the bye falls, and how many games sit either side of it. How many consecutive road games appear, and whether any run reaches three. Where the short weeks are, and whether one lands on the back of a long trip. How many time zones the club crosses and when. Which fixtures are in national windows, and how the week after each of them looks.

Check the home and away split and whether this is a nine-game or eight-game home season. That follows from which conference hosts the seventeenth game, and it flips next year. It changes the shape of a season more than most people account for.

Mark the flex-eligible weeks and treat those dates as provisional. The fixture will not change. The day and the kickoff time might, and if your club is having a good year they probably will.

Do that and you will know, in about ten minutes, which parts of the year ahead were decided by a formula in January and which parts somebody actually chose. It is a better basis for an argument than a strength-of-schedule ranking built on last season's records, and it is roughly the same amount of work. Most of what makes this sport legible is knowing which things were decided and which were merely arranged, and the schedule is the clearest example on the calendar. The fixtures were arithmetic. Only the dates were a decision.

You will also, before a single game is played, have noticed something the analysis lists never mention: that the warm-up fixtures that precede all of this are the only games on the calendar nobody had to solve a puzzle to arrange.