What the model aims for.

What the model aims for

The weighted rule (show definition)weighted ruleThe AFL's way of sharing out double-ups by ladder position, so strong clubs mostly play strong clubs twice.Read more in the glossary allows up to four double-ups (show definition)double-upA pair of clubs that meet twice in the season. Each club has six.Read more in the glossary from a club’s own ladder group. Within that limit, the model prefers no more than three.

Why

The AFL (show definition)AFLAustralian Football League.Read more in the glossary's weighted rule exists to even out the difficulty of each club's draw. The preference nudges the model toward the AFL's usual pattern rather than the edge of what is allowed.

The trade-off

Pushing harder on this preference leaves less room for the venue rules. When it was made a must-follow rule together with the Victorian venue caps, no fixture satisfied both. It returned as a preference on 18 September, tried at two strengths; the stronger one was kept in both calendars.

See it in the fixture

Status. Declared preference in force from the 18 September release.

What the model aims for

The model prefers evenly spread byes (show definition)byeA round in which a club does not play.Read more in the glossary and as many games in bye rounds as the rules allow. It also prefers rematches further apart than the six-round minimum.

Why

Even bye rounds keep every weekend worth watching; well-spaced rematches feel right to fans.

The trade-off

Spacing rematches further apart leaves fewer ways to fit the fixed occasions and byes. In early versions the model's smallest rematch gap was twelve rounds, twice the minimum.

See it in the fixture

Status. Declared preferences. The public release does not say how strongly each counts.

What the model aims for

The model can prefer fixtures where the most appealing match-ups are spread across rounds rather than bunched.

Why

A season with all its blockbusters in April and none in July is a worse season to follow.

The trade-off

Spreading big games can conflict with the fixed occasions, which cluster several of them around Anzac Day and the King's Birthday weekend.

See it in the fixture

Status. A declared preference, set to have no effect by default. No release note records it being switched on. Treat it as off unless a release says otherwise.

What the model aims for

When choosing Thursday and Friday night games, the model prefers those ranked highest by its declared appeal policy.

Why

Thursday and Friday nights are the biggest television slots. The AFL puts its best games there, and the model follows the same policy.

The trade-off

Games ranked lower get fewer prime-time nights. In the current release, Essendon has no Friday-night games on either calendar. The release names this outcome instead of changing the scores to hide it.

Limits it must follow

The model must obey each club’s limits on weekday games, seasonal spread and recent weekday range. These are must-follow rules. The preference chooses only among fixtures that satisfy them.

See it in the fixture

Status. In force from the 30 September release. The release calls it the “declared prime-time objective at Step 3”.

What the model aims for

The model first improves the round with the lowest combined prime-time appeal ranking as much as it can. Then it maximises the season total.

Why

Without this, the model could stack every big game early and leave late-season rounds thin. Every round should have something worth watching.

The trade-off

The season total of appeal is a little lower than it could be, because some strong games are held back for otherwise weak rounds.

See it in the fixture

Status. Declared policy in the 30 September release notes ("No fading season").

What the model aims for

Each club has an appeal score that weights recent seasons most heavily. It draws on how many prime-time games the AFL gave the club, its ladder form, and popularity tiers drawn by hand. A game's appeal combines its two clubs' scores, with a bonus for derbies.

Why

The model needs some way to say which games are bigger. This is a declared policy, chosen by the project owner, not a measurement of anything.

The trade-off

The score is only as good as its inputs. A club with poor recent form ranks low even if it is well supported. The tiers were informed by membership and attendance data but are not bound by it.

See it in the fixture

Status. The scores themselves are not published on this site. The method (recency-weighted club score with a derby bonus) is declared in the season settings and was adopted on 28 and 29 September 2026.

What the model aims for

The model counts each club's interstate games and prefers fixtures where that count is spread fairly between clubs. It does not measure kilometres.

Why

Travel is tiring and costly. The AFL tries to keep it fair, and fans notice when one club flies every second week.

The trade-off

Fewer trips for one club means more for another. Every club must still visit Western Australia and Victoria (must-follow rules), and Gather Round (show definition)Gather RoundThe round in which every game is played in one host city. Every club appears once.Read more in the glossary sends everyone to one city.

See it in the fixture

Status. A declared preference. The model counts interstate games; it does not use the catalogue’s distance-based travel preference. The public release does not say how strongly this preference counts.

What the model aims for

The model prefers fixtures where no club has long runs of home games or long runs of away games.

Why

The AFL prefers at most two home games and two away games in any three weeks. It acknowledges it cannot always achieve this.

The trade-off

Shortening one club’s streak can lengthen another’s. The model allows no more than three rounds without a listed home game. Within that limit, it prefers shorter streaks.

See it in the fixture

Status. A declared preference. The release reports departures from the AFL’s two-in-three preference without failing the fixture. The public release does not say how strongly this preference counts.

What the model aims for

Must-follow rules settle most ground choices. These include contracts, sold games (show definition)sold gameA home game played at another ground under a deal, such as Hawthorn in Launceston or Gold Coast in Darwin.Read more in the glossary, ground minimums and limits, and the rule for big Victorian games at the MCG (show definition)MCGMelbourne Cricket Ground.Read more in the glossary. Within those limits, the model considers the travel and home-streak preferences above.

Why

Most of what decides a ground in real life is a contract, not a preference.

The trade-off

The AFL also considers expected crowds when choosing grounds: a smaller crowd may suit a smaller ground. This model does not make crowd-based ground choices. A proposed attendance model failed its test and was not adopted.

See it in the fixture

Status. The public release records no separate preference for ground choice.

What the model aims for

The model first finds TV slots that maximise its free-to-air (show definition)free-to-airTelevision anyone can watch without a subscription.Read more in the glossary audience prediction. That prediction comes from a model fitted to published TV ratings. It then searches for the highest combined match appeal, while keeping the prediction within two per cent of the first result.

Why

Broadcasters want more viewers, while fans and the AFL want strong games in the best slots. The second pass balances these aims within a set limit.

The trade-off

Some games lack a prediction, most often South Australian and Western Australian games. They keep a reference slot (show definition)reference slotA slot a game keeps when the viewing model has no basis for predicting its alternatives.Read more in the glossary and do not enter the slot search. This explains Port Adelaide’s few Saturday nights on the Wednesday-opening calendar: a data gap rather than a scheduling choice.

Limits it must follow

Unmeasured audiences are treated as unknown, never zero. Pinned games (show definition)pinned gameA game the audience model cannot score, so it keeps a reference slot and is not optimised.Read more in the glossary are capped at one in five of the season. Pay-television-only games are counted separately and never added into the free-to-air figure.

See it in the fixture

Status. In force from the 30 September release. Predictions use model units, not measured viewer counts, and are not published here. Earlier releases found slots that followed the rules, without preferring particular games for particular slots.

What the model aims for

After building the fixture, a realism report (show definition)realism reportA comparison with recent AFL fixtures, checking each club’s weekdays, slots, home grounds and longest run without a home game.Read more in the glossary compares it with the AFL’s 2023 to 2026 fixtures. It checks each club’s weekdays, slots, home grounds and longest run without a home game. Results far outside a club’s history get a severe flag (show definition)severe flagA realism-report result far outside a club's 2023 to 2026 range. Needs a stated reason before release.Read more in the glossary. A release needs a stated reason for every severe flag before publication.

Why

A fixture can obey every written rule and still not look like a fixture. One earlier version gave Adelaide 21 or 22 Thursday nights (show definition)Thursday nightThe week's first game. The 2027 model has 24 of them, plus one Wednesday night on the 17 March calendar.Read more in the glossary without breaking a single rule.

The trade-off

Realism pulls toward the past. A deliberate policy change (for example, more prime time (show definition)prime timeThursday and Friday nights, the biggest television slots.Read more in the glossary for a rising club) will show up as a severe flag and needs an explicit reason.

See it in the fixture

Status. A release check, rather than a preference used to build the fixture. Some measures have since become must-follow rules. The current release is “clear”: every severe flag has a stated reason.

What the model aims for

A release can build several fixtures using different random starting points, preference strengths (show definition)preference strengthHow much an aim counts when the model balances it against other aims.Read more in the glossary or calendars. It checks and compares them on several measures. It keeps those that no other fixture beats on every measure at once, then names a headline version.

Why

There is rarely one best fixture. Showing the alternatives and comparison measures helps readers understand the choice.

The trade-off

Keeping several candidates means the headline is a judgement call among them. The release says which was chosen and on what comparison.

See it in the fixture

Status. The 18 September release compared four fixtures: two calendars at two preference strengths. It chose the stronger preference for both calendars. The release notes do not describe the keep-only-the-unbeaten step. That comes from the generator’s description of how it selects alternatives.

What the model aims for

For Steps 1 to 4, the program must prove no better answer exists within that step’s rules before publication. Step 5 records its result and how close it came to that proof.

Why

A step that reaches its time limit (show definition)time limitHow long each stage may run. The 30 September release allowed ten minutes per stage.Read more in the glossary might give a different answer next time. Repeatable results let others check the work.

The trade-off

Proof takes time. The 30 September release allowed ten minutes per step. A run that cannot prove Steps 1 to 4 within that time cannot be published.

See it in the fixture

Status. "Proven best" means best for that stage, given everything already decided by the stages before it. It never means the best possible season overall.