Rules
The short version: everyone answers the same prompt, the Judge throws out what does not fit, search-interest data scores what is left, and the highest total after the last round wins. This page is the long version, with the exact numbers the game uses. The How to play page walks through a single round first if you have never played.
Scoring
- Search interest, 0–100. Every valid answer is measured against real search-interest data over the scoring window, 12 months by default. The round's answers are scored together as one batch, so the hottest point in the batch is 100 and everything else sits relative to it.
- Points from score. A score of 25 or more is banked as it is. Below 25 the game applies a boost curve, 5 × √score, so a quiet but real answer still earns meaningful points and no round goes dead.
- ⚡ Peak, +20. The player whose term owns the single hottest week on the round's chart earns a +20 bonus. The newest data point on the chart never counts as the peak, because it is still provisional.
- The Judge and off-topic answers. An automated judge checks each answer against the prompt before scoring. Categories are checked strictly for membership (an apple is not a vegetable); complete-the-search prompts are checked leniently for whether the finished phrase is a plausible real search. A rejected answer scores 0, draws no chart line and spends no data. You may pre-check an answer with Ask the Judge before locking it; a pre-check approval is binding and cannot be overturned at scoring. Free players get one check per 180 seconds, Party Pass holders unlimited.
- No data. If none of a round's answers exists in the search data, the answers that come back identical are scored as no data: 0 points, no line. A whole batch of zeros is treated as a broken measurement instead: the round is voided and replayed in the same slot with a fresh prompt, and any auction bid paid that round is refunded.
- Same answers. What happens when two players type the same term is a room setting. First Dibs (the default): the first player to lock the term keeps it and later players are told to try another. Split: the term is scored once and its points divide evenly among everyone who typed it. Auction (host opt-in): the holder and the challenger each place one sealed bid of 2 to 30 points within 12 seconds; the higher bid wins the term and pays the bid to the loser, ties keep it with the holder, and whoever ends up without an answer is guaranteed at least 15 seconds to type a new one. Word-order twins such as “cat house” and “house cat” count as the same term.
- Word rules. Within a party, a term played in an earlier round is banned for the rest of the night, and each content word of it is banned for the next 30 rounds. An answer made only of function words (“how to”, “best”, “free”) is refused before scoring, as is an answer that repeats the prompt's own words. The live bans are listed in the room under Used words.
Special rounds
Each round after the first has a small chance of a twist, announced before it starts: Reverse (the lowest real search volume wins; real answers score a spread from 30 to 90 with the quietest at 90, and zero volume still banks 0), Double (every point counts twice) and Trending today (scored on the last 24 hours instead of 12 months, falling back to 7 days and then 12 months if the fresh data is too thin). The final round is always ×2, and neither it nor the round before it ever rolls a random twist. The host can switch rolled twists off; the final round's ×2 stays regardless.
Seating
- Solo seats is the regular party room: up to 5 players, each guessing for themselves. Rooms are joined with a 4-letter code, a QR code or a link, with any name; names are unique in a room, and a second “Tom” becomes “Tom the Greatest”.
- Hive mind is the room for big groups: 6 to 100 players in 2 to 5 teams of up to 20. A team round is 60 seconds in two halves: 30 seconds for every member to suggest a term, then 30 seconds to vote; the most-voted suggestion is the team's answer, a tie goes to the Judge's coin, and two teams on the same term flip for it, with the loser falling back to its second choice or a 15-second re-pick. Every member banks the team's points, and the player who suggested a round-winning term earns an MVP credit.
- Trendmaster makes the host a human referee who does not play: the automated judge is off, the host approves or rejects every answer and may award −100 to +100 bonus points. Creators mode has the players write the prompts, each sitting out the round built on their own question.
- Round settings (host, lobby or between games only): 1–50 rounds per game, default 5, or 0 to play until the challenge pool runs out; 0–600 seconds per round, default 90, where 0 means the host closes the round by hand. Every round opens with a 3-2-1 countdown on every screen, so the host gets no head start.
- Hosting. If the host leaves or goes quiet for 20 seconds while others are active, the room hands the role to the longest-present player. A host can remove a player only in the lobby or between games, never mid-game.
Play solo and ghosts
- Play solo joins a queue; the game guarantees a match by seating ghosts in any empty seats. Rounds are 60 seconds, the game paces itself, and same answers are pinned to First Dibs.
- A ghost replays a real answer a real player gave to the same prompt in the past, with human-like timing, under a made-up name such as Ghost Otter.
- Fairness rules. A ghost never replays an answer the game would refuse a human, such as “best” or “how to”. When the archive holds more answers than there are ghost seats, the strongest third is held back, so a ghost rarely takes the top of a round a human had a real shot at. A solo game never shows a player a prompt they answered in their last 50 rounds.
- A game with only one human at the table is unranked; the game tells you when another real player opens a queue, and switching to it costs nothing.
- Play solo is where the game's only ad runs, during the wait in the queue and never during a game.
Parties and rematch
- Play again starts the next game in the same room: the roster, the settings and the word bans carry over, and no prompt repeats within the party.
- Between games the room keeps a party table, a cumulative leaderboard for the night with each player's game wins, plus awards built from what actually happened: Most peaks ⚡, Best steal, The Judge's least favorite and Comeback of the night.
- The winner's slogan: the winner of a game gets one 30-second window to type a victory line that the whole room sees.
- Leaving is immediate and your seat and points stay at the table; come back with the same name and room code and the seat is yours again. A host leaving mid-game is asked once, because leaving hands the room over.
- Players vote on the next round's topic while the reveal and standings are up: a ballot of two challenge packs the room has not exhausted, most votes wins, ties go to chance.
Ranked play
- Signed-in players earn Rank Points (RP) from every ranked match. Placement is worth 50 / 30 / 15 / 5 / 0 points for 1st to 5th and below, minus a tier cost (0, 0, 10, 20, 30, 40 by tier), scaled by the game's length (×1.0 at 5 rounds up to ×3.0 at 30) and by how many humans competed (×1.0 with four or more, ×0.6 with exactly three), plus a participation bonus of +5 to +20.
- A match is unranked when two or fewer humans competed, when the prompt pool contains custom or player-written challenges, or when a round had to fall back to demo scoring because the data provider was unreachable.
- Tiers: Human in Training (0 RP), Mildly Curious Human (300), 3 AM Human (600), Human Trendsetters (1080), Human Mind Reader (1680) and Knows Why Human Ask (2400). Divisions count up, I to III. A loss never drops you below the floor of the major tier you started the match in, and a fresh promotion grants two demotion shields.
- Quitting a ranked match in progress costs −35 RP. Top-tier players who play no ranked match for 7 days lose RP daily until they drop a tier. Seasons last 90 days and soft-reset every tier.
- Hive mind games are ranked at half stakes (25 / 15 / 8 / 3 / 0) when at least two teams have three humans each, plus +5 RP per MVP round, capped at +20.
Fairness principles
The server is the referee; phones only show what it decided. Every score traces to a live measurement, never a stored or invented number, and broken data voids a round rather than shipping a fake result. A pre-check approval binds the final judge. Nobody is stranded: an auction loser gets a guaranteed window to re-answer, voids refund bids, and a round never closes over a live auction. Settings never change under a live round.