Modes
Two families of modes, and what each one gives.
Face Down is played live, against another player, and its outcome depends on reading them. The Trends ask what will be read next, against what was expected, and that answer doesn't exist when you pick. This difference decides what each mode can give.
Why two families
A live duel is won by reading another person, and a script that reads a predictable player can do that too. So the duel is there for the fun of playing: it gives INK, IQ, cosmetics and titles, and boosters for the top of its weekly standing.
A prediction mode asks about views that haven't happened yet, judged against an expected figure published for everyone, so there's nothing to fetch, copy or automate. That's why most of the standings are about prediction.
Every reward on this page is in the game: INK, boosters and season cosmetics. None of it can be bought, and none of it converts into money. What each standing gives →
Real-time modes
Face Down: the duel
Both players get the same nine cards, and nine trophies (famous cards worth their Attention) are revealed one at a time. Each round, both players put a card face down at the same moment, with thirty seconds on the clock. The cards turn over, the higher Attention takes the trophy, and both cards leave the hands. The bigger total of trophies wins. Both players see everything, and the game is about reading the other player. Duels are matched on IQ, like trophies in a ladder game. You never face someone whose hidden level is far from yours, the cards you own play no part in matching, and you're never matched against a friend. A win is worth 28 to 35 IQ and a loss costs 28 to 35, depending on how strong the opponent is.
A matched duel gives INK (25 for a win, 5 for a loss) and a streak bonus in INK from three wins in a row, with titles at five and ten. It moves your IQ, the prestige ladder of ten tiers. Each tier has a floor that only two losses in a row at that tier can take away (tier 1 and tier 10 never drop, and what a tier gave you stays yours), and there's a small reward every 75 IQ between them. Above the tenth is the Legend league, where IQ moves both ways and each monthly season halves what you hold above tier 10. Monthly seasons give dated titles that never come back. A win drops one victory booster into a free slot, at most ten a day.
It also has a weekly standing. The first five matched duels of each game day count, won or lost, and the week's points (Monday 12:00 to Monday 12:00 UTC) decide it: the top tenth get a Chapter, and the top hundredth a Codex instead and, the first time in the season, the title "Duel elite · All Hallows 2026". While you wait, you can choose to play The Archivist instead (it's never forced on you). Its duels give INK only, 150 a day at most: no IQ, no booster, nothing counted.
Friend duel: 4 + 4 + 1
This is Face Down with a friend, on one hand shared by both: four cards you chose, four your friend chose, and one drawn at random. Each of you sees the other's four at the start, and since you both play the same nine, a bigger collection wins nothing. It gives INK only, on your first five friend duels of the day and on five duels per pair of friends a week. It never gives IQ, a booster or a place in a standing.
The open duel, Ranked Draft and challenges by link are paused.
Prediction modes
Trend Call: every day at 12:00 UTC
Five pairs a day, the same five for everyone, drawn from the whole pool: cards and comparison reserve, people included. The question is which one will beat its expected views by more?
The two subjects of a pair are matched on their seven-day daily average, within 5% of each other. Each subject comes with an expected figure: the views it should get between 12:00 and 18:00 UTC, computed from the hours already published, its habits, its size and its trend. Picking the bigger name doesn't help, because the expected already knows it. The winner is the subject furthest above its own expected, so the most-read one doesn't always win.
- The window: the day's five pairs appear the day before at 12:00 UTC, and you have 24 hours to pick, until 12:00 UTC on the day itself. One pick per pair per account.
- The measure: the pair is judged on that day's views between 12:00 and 18:00 UTC, read from the hourly files Wikimedia publishes. While picks are open, each subject shows a provisional expected and its views "so far"; at the close, the server publishes the frozen expected of each subject, the same for everyone.
- The result comes out the next day at 12:00 UTC: "Beat it" or "Missed" for each subject, with the gap to its expected, at the same moment the following day's pairs open.
- Streaks give boosters: 3 correct in a row → a Page, 7 → a Chapter, 12 → a Codex, waiting on your shelf with no timer. A wrong pick resets the streak. Each correct pick also gives 10 INK.
- The week is counted at 25 out of 35. One point per correct pick, over the whole week, not just the days you played. A week at 25 correct or more out of its 35 pairs gives a Codex and, the first time in the season, the title "Trend reader · All Hallows 2026". It is settled on the following Monday, after 12:00 UTC.
- A perfect day gives 25 more INK: the day's five pairs all counted and all right.
- A death, a tie or a missing hour neutralises the pair: no point, no effect on the streak.
The Open League: every week
Once a week each player picks three subjects from a common list of 16. The list doesn't come from anyone's collection, so owning more gives no advantage. Entry is free, one entry per account.
The list appears Monday at 12:00 UTC and picks close Sunday at 23:59 UTC. The measured week runs the following Monday to Sunday, and the result comes out the Monday after. You always pick before the first hour of the week being measured, never during it.
The question is which three will grow more than expected? Every subject on the list comes with an expected growth, published for everyone and frozen when picks close. At resolution the 16 subjects are ranked on their real growth divided by their expected growth. First place is worth 16 points and last place 1, so three picks give a score between 6 and 45. Lists have had this size since 19 October 2026; earlier weeks keep their size (15 subjects on 12 October, 30 subjects and five picks on 5 October) and count in the standing at the same scale.
Three random picks are worth 25.5 ± 7.4, and that luck line is published on the standings. Skill has to beat luck by a clear margin, or the rest of the league means nothing. Since the expected already accounts for a subject's size and trend, the skill is knowing the calendar: releases, fixtures, anniversaries, and what nobody has priced in yet.
- The list is drawn by the server across the whole pool, people included. Each subject needs at least 2,000 views a day over the previous 28 days, can't be flagged, and can't have been used in the last eight weeks. The list is published with its week number and each subject's expected growth, and everyone's picks are revealed when picks close.
- A subject who dies during the week is neutralised: the pick is worth the median position, neither gain nor loss.
- The standing rolls over four weeks and you must have played at least three of them. On equal points, the player whose picks beat their expected growth by more goes first (the average of real ÷ expected over their picks); players still level share the same rank. Each league gives 20 INK for every week you play. Each Monday, the top tenth of the four-week standing get a Chapter and 100 INK, and the top hundredth a Codex instead of the Chapter, with the same 100 INK.
- Everything can be checked: the list, the expected figures, the picks, the final figures, and the interface calls that produced them. How expected is computed →
Themed leagues
Themed leagues work like the Open League (three picks from a common list, each subject with its expected growth, score by position, deaths neutralised), with the list limited to one theme. They run alongside the Open League every week and never replace it.
A themed list has 16 subjects, the same size as the open one, with a floor of 1,000 views a day and no subject repeated within eight weeks. Every league has its own standing and its own rewards, and they are never added together. A score is a position inside its own list and never leaves it, so the size of a list changes nothing about any reward.
A theme opens only if enough subjects hold its floor in every month of the year, and the test is repeated every quarter. Games is suspended on that test; Basketball and American football stay closed as too thin.
| League | What is in the list |
|---|---|
| Football League | Players, managers, clubs, competitions, stadiums |
| Screen League | Cinema and television: actors, directors, films, series |
| Music League | Singers, rappers, musicians, composers and bands |
| Games League | Video games, studios, characters, franchises. Suspended since 24 September 2026: too few subjects hold the floor all year, so it publishes no list until a quarterly test says otherwise. |
| Minds League | Sciences and ideas: writers, scientists, philosophers, concepts |
| Politics League | Politicians, elections, parties, parliaments and institutions |
| Tech & business League | Companies, brands, founders and executives, platforms, software |
The politics league has three rules of its own, because a weekly list of political figures is the one theme that can read like a scoreboard of people. Nothing from it (no list, no result, no standing) is ever published on the project's accounts. It has no list at all during the week of a major national election, following a calendar published in advance. And nothing is ever featured because someone died.
There is no elections league, war league or trial league. Those are themes of events, and they stay closed. Politicians remain ordinary cards in the Open League, which draws from the whole pool.
The Football League never asks who will win a match, what the score will be or whether a transfer happens. It asks which articles will be read most this week, which depends on football being in the news, whatever its results.
The short modes
Up or Down: once a week
Every Monday at 12:00 UTC, with the leagues, comes a list of 30 subjects, the same for everyone. For each one, you answer a single question: better or worse than expected? Each subject shows its expected for next week (next week's views over this week's), published for everyone and frozen when answers close. Answers close Sunday at 23:59 UTC, and the result comes out two Mondays later: the 15 subjects furthest above their expected are "better", the other 15 are "worse", with no neutral zone. A subject who dies leaves the list, and a subject left blank counts as wrong. It gives INK and titles: 5 INK per right answer, 25 more if all of them are right. Each Monday a result comes out, the top tenth of a rolling four-week standing, with at least three weeks played, get a Chapter.
Timeline and Rank Them, two more short modes, are planned.
What each mode gives
| Mode | Rhythm | Gives | Standing rewards |
|---|---|---|---|
| Face Down | Any time; five counted duels a day; monthly season | IQ, INK, streak bonus, titles, victory boosters | Weekly: a Chapter for the top tenth, a Codex for the top hundredth |
| Trend Call | Daily, 12:00 UTC | INK, streak boosters, 25 INK for a perfect day | A Codex for a week at 25 correct out of 35 |
| The Open League | Weekly | 20 INK per week played | Four weeks: a Chapter and 100 INK for the top tenth, a Codex and 100 INK for the top hundredth |
| Themed leagues | Weekly, all at once | 20 INK per week played, in each league | The same, for each league apart |
| Up or Down | Weekly | INK, titles | Four weeks: a Chapter for the top tenth |