๐๏ธ A town built from real people
Your character moves into a town of 1,000 made-up but realistic adults, built from real US data: who they are and what they do, who's single or married, who's gay, straight or bi, and where people spend every half hour of a normal day.
โฑ๏ธ A month, half an hour at a time
The game plays out 30 days in half-hour steps. Everyone goes about their life: work, the gym, the shops, dinner, a bar, home. When people end up in the same place, they might start talking.
๐ First impressions
When two people meet, each one quietly decides: would I see them again? That decision comes from a real experiment, where thousands of strangers went on 4-minute speed dates and said yes or no. Looks count the most, then being fun, then having things in common. Strangers misjudge each other a little, so luck plays a part.
๐ท From spark to sleepover
- Both say yes and it's late at a bar: you might go home together that night.
- Otherwise you set up a date sometime in the next week, and a date can end at someone's place.
- Book two dates for the same night and someone gets stood up.
- On nights in, you swipe on the apps at 10pm. Profiles only show looks. A match becomes a date.
- People gossip: sleep with someone's friend and that friend might cool on you.
๐๏ธ What your stats do
You have 33 points to share out, and each extra point is worth a bit less than the one before, so piling everything into one stat rarely wins.
- ๐ Looks: the apps and every first impression.
- ๐ Fun: first impressions, and the odds a night out ends at someone's place.
- ๐ Sincerity: the odds a date ends at someone's place. (Strangers barely notice it at first.)
- ๐ง Smarts: conversations click when you meet in person.
- ๐ Ambition: you book dates on your next free night, with no double-booking.
๐
Your week
- Nights out: each one is a bar crawl from 7pm to midnight. It's a night off the apps.
- On the apps: a steady stream of dates, as long as you have free nights for them.
- Keeping it casual means one night and on to the next. Love means you might settle down with someone, which is sweet, but your count stops climbing.
- Your interests help you click with people who share them, in person.
Apps-only players, bar-hoppers and a mix of both can all do well, but they want different stats.
โ๏ธ Your description
Before your month starts, an AI called Jev reads what you wrote and asks, for each person in town who could be into you: would they want a date with this character? Their answers decide who likes you, and the town's overall verdict decides how many: a charming description can lift your score by a third or more, and an off-putting one can halve it. Specific and charming beats bragging. "A perfect 10" gets you nothing, and nor do instructions aimed at the AI. Race, ethnicity, body size and disability are ignored.
๐ณ๏ธโ๐ Gay characters
A town of 1,000 has only a handful of gay singles, so for gay characters the apps also reach singles from out of town. That makes the apps their best bet, and every kind of character has a real shot at the top.
๐ค Technical details for nerds
- Attraction: two logistic regressions (women choosing, men choosing) fitted on ~3,400 speed-dating decisions each. P(yes) = ฯ(ฮฒโ + ฮฃ ฮฒแตขxแตข) over the chooser's ratings and the age gap. AUC 0.81 / 0.84. Looks carry the most weight, then fun and shared interests.
- Noise: ratings are seen with Gaussian noise (sd 1 point in person, 1.5 on the apps). Race is held at the dataset average, so it never changes the odds.
- Extra logit terms: age-gap penalty, gossip, Jev's offsets and the player's charm bonus.
- Schedules: ATUS diaries give the odds of being at each place every half hour, by job, age and day of the week.
- Breakups: a daily hazard from HCMST, high in a couple's first six months and lower after.
- Calibration: ten knobs fitted by coordinate search so a simulated year matches GSS partner counts by sex and age, checked on seeds the fit never saw.
- Determinism: a seeded RNG gives the same month on the server and in your browser.
- Data sources: General Social Survey 2022โ24 (NORC), Fisman & Iyengar speed-dating experiment (Columbia), How Couples Meet and Stay Together (Stanford), American Time Use Survey 2024 (BLS), Gallup, NVIDIA Nemotron-Personas-USA.
Built by @mizanxali