A charity awareness piece · Toronto

Homeless in Toronto

The dice are fair. The board is not.

Homeless in Toronto is a 3D browser board game about this city's housing arithmetic. You play a tenant. You always lose. That is not a difficulty setting — it is the argument, which is that the arithmetic of Toronto housing does not close for a person on Ontario Works no matter how the dice fall.

Play the piece

This is the text version: the same argument and the same sources, with no WebGL, no animation and nothing to install.

What it is

32 tiles in a ring. You roll two dice, walk, and a receipt tells you what the tile cost you and — where a real cited figure explains it — what the system it belongs to actually does. You go round 3 times. Each lap is a month, and each month your income falls while the rent stays exactly where it is.

The engine, in 3 rows
MonthIncomeRentResult
1$2,210 — part-time, 29 h/week at minimum wage$1,491$719 left for everything else
2$1,105 — hours cut in half$1,491$386 short; arrears begin
3$733 — Ontario Works, all of it$1,491$758 short before food

The rent never moves. That is the point: the game never inflates the rent to make its case, it simply lets the income collapse under a real number. There is no fourth lap, no score, no leaderboard, no streak and no win state.

How the data works

Every statistic in the game is real, cited, and lives in a JSON file you can read in the repository. The rule the whole project is built on is that no number about homelessness, rent, deaths, shelter capacity, food waste or waitlists is ever generated, estimated, inferred, rounded or plausibly filled in. Where a figure could not be sourced, the record is marked verified: false, excluded from play by the loader, and reported as a warning when the game boots.

This page is generated from those same records by tools/build-about.js. It is not a second copy maintained by hand: every figure and link below is read out of content/ at build time, so the page cannot quietly disagree with the game.

Records currently excluded, and kept visible

One number in the game is not a statistic, and says so

The game charges a fixed monthly non-rent budget line of $690 for food, transit, phone, laundry. That figure is a game mechanic, not a measurement, and it is labelled as exactly that everywhere it appears: the receipt's own delta row reads "game budget", and every rent receipt states in words that it is "a fixed game budget line; not a real-world statistic". It exists because "$719 left for everything else" needs an everything else, and because without it a lucky run could bank the surplus into survival. The real number sits beside it: Food bank clients report a median of $8.83 a day left after paying for housing. — Daily Bread Food Bank, Who's Hungry 2025.

Where the money goes

Nowhere near this project. Homeless in Toronto collects nothing, processes nothing and takes no cut. Every donation link is an outbound link to the organisation's own site. Where an organisation's own donation page could be confirmed by following a Donate link on its own site, the button goes straight there; where it could not be confirmed, the button says "Visit" instead of "Donate" and goes to the front page, because guessing a donation URL is not something this project does.

If you need help right now

Whose game this was

Lizzie Magie patented The Landlord's Game around 1904 as a teaching tool, built to demonstrate that land monopoly impoverishes people. The game that made a fortune from her idea inverted the lesson and taught a generation that winning meant owning everything. This piece restores her intent.

Homeless in Toronto restores her intent. It uses none of that game's name, characters or trade dress, and it has no win state to invert.

Every source in the game

Verbatim, with links, straight from the records. A citation with no public URL says so rather than pretending to have one.

The figures the engine uses

The system deck

Mutual aid

The ending

highlight the failure of multiple systems, such as affordable housing, health care, mental health, income support, and the justice system… People are left with nowhere to turn, and the shelter system or staying outdoors become the last resort.

— City of Toronto, 2024 Street Needs Assessment

Questions

Can you actually win?
No. There is no win state and no code path that produces one. The game verifies this at boot: it bounds the most generous outcome the content permits and confirms that bound still falls short of the rent.
Is this a real person?
No. The game assigns a first name, an age and a line of work so the arithmetic has somebody to happen to, and labels it a composite, not a real person. There is no rendered face, no photograph, and no statistic in the game is ever attached to the individual.
Why does the player always lose?
Because a game you can win would be an argument that the situation is winnable. The Ontario Works shelter allowance is $390 a month and the average purpose-built bachelor apartment in Toronto rents for $1,491. That is a gap of $1,101 every month, before food. No sequence of good decisions closes it, so the game does not offer one.
Does the game blame anyone for being homeless?
It is built not to. The player is always the tenant and never the landlord. Health and substance use enter only as consequence — untreated pain, no doctor, a supply nobody can verify — never as cause or character flaw, and never attached to a neighbourhood. A specific set of stigmatizing labels for people experiencing homelessness is banned outright from the entire project, in content, code and comments alike.
Where do donations go?
Directly to the organisations, on their own sites. This project has no payment processing of any kind and could not take a cut if it wanted to.
Is any of the artwork AI-generated or downloaded?
The board, the city and every texture on them are generated in code and drawn to a canvas when the page loads — the concrete, the brick, the lit windows, the dice pips, the snow, the street signs, the flag and the favicon. The painted backdrop, the intro frames, the two character portraits and the card illustrations are illustrated artwork shipped with the build. There are no downloaded models and no photographs anywhere in it.
How long does it take?
Six to ten minutes to play, and it works on a phone. It is fully playable with a keyboard alone, and fully playable with animation switched off.