Watchfires — a stacking board game of beacons for 2–4, all printed

A stacking board game for 2–4 players where everything on the table is printed: one hexagonal plate, 36 rings, 4 hills. No dice, no cards, no text.

▶ Watch it explained and played — 10 minutes: https://youtu.be/ThitDa0DtpY

Nineteen posts on the plate; a post holds three rings. The outline of a ring — circle, square, triangle, hexagon — says whose it is, so any four spools work. Placing a ring costs as many steps round the clock (the loop of posts on the rim) as the tower is tall afterwards, and the player who has spent the least always moves next. That one rule is the whole turn order, and it is why the seat you start in makes no measurable difference. At the end every ring is a watchman on its floor, counting the towers it can see along the six lines. Highest total wins.

Six STLs, 41 pieces, about 136 g, no supports on anything; an optional lid (66 g) covers the loaded plate — a dust cover, not a travel latch: it does not grip and it does not hold. A ten-minute fit coupon (one post, three ring bores) goes first, and plate_2c.3mf prints the lattice, numerals and sun in a second colour for about a gram of purge — it opens as one object already split into two parts with the filaments assigned, so there is nothing to click (plate_2c.stl is the same geometry for slicers that take no 3MF). The plate (173 × 150 mm) prints in about two hours; each plate of nine rings in 25 minutes. The plate stores the whole set on its own posts.

Rules (one page, with a typeset PDF and nine machine translations), a print guide and the measured rating — luck vs skill, seat balance at 2, 3 and 4 players, game length — are included.

Print it

1. The coupon, first — ten minutes. One post exactly as the plate prints it, and three rings bored 5.4 / 5.6 / 5.8 mm (labelled 54 / 56 / 58). Whichever ring drops on and lifts off without rocking is your tolerance; if that is not the 5.6 middle one, set ring_bore in the .scad before committing two hours to the plate.

2. The plate — about two hours, 94 g, no supports. 173 x 150 mm, so it fits any 180 mm bed. Print plate.stl in one colour, or plate_2c.3mf for two: it opens as ONE object already split into plate_body and plate_ink with the filaments assigned, so there is nothing to click. plate_2c.stl is the same two solids for a slicer that takes no 3MF — Import, right-click, Split to Parts. About a gram of purge either way; the ink is the top three layers only.

3. Nine pieces per player, 25 minutes a plate. rings_circle, rings_square, rings_triangle, rings_hexagon — one plate each, whatever colours you have. The OUTLINE is what says whose a ring is, so four spools of the same grey is a legal set.

The ninth piece on each plate is not a spare. Eight rings build; the ninth is that player's marker on the clock, and it has rays cut clean through so a piece set down beside the board cannot be mistaken for a ring. Same outline, same 4 mm floor.

4. Four hills, one gram each. One per player, a one-floor rocky base. It goes down first and everything on top of it stands a floor higher.

5. The lid is optional — 75 g. A hex cap that sits over the loaded plate with the pieces still on their posts. lid_2c.stl is the two-colour version (Split to Parts: lid_body + lid_ink); it prints face-down and needs no supports.

The lid is a cover, not a catch — printed and tested. It drops over the loaded plate and hides it, but it does not grip the rim and it does not hold: lift the set by the lid and the lid comes away, and tilt it and the pieces move. Print it if you want the box to look finished on a shelf. Do not count on it to carry a loaded plate anywhere.

6. There is no assembly. Nothing glues, nothing snaps, no hardware. Rings drop onto posts and lift off again — that is the whole game. The plate stores the entire set on its own posts, under the lid.

STARS:BEGIN

At a glance

There are no dice. The hills are the whole opening; everything after is open information. Playing the rules properly beats random play 200 of 200; looking one placement ahead beats playing properly 88 % of the time; searching harder beats that 67 / 33. Replaying the same hills with one seat swapped good→random changes the winner in 58 % of games. Start position is worth nothing measurable at two or three players (n = 800 each) nor at four (24 / 22.5 / 25 / 24 over 1,800 games) — the eighteen-step clock was chosen for exactly that. Measured L = 0.60 at two players, 0.79 at four.

The dials

More stars means more of the named thing — never "better". Every row names both ends, so nothing here needs a legend.

STARS:END