Burrow

Sixty seconds of moles. Tap them. Leave the bombs alone — one costs you five seconds. The crowned ones are worth three, and do not wait around.

Open index.html. No build step, no dependencies, no images.

burrow/
  index.html      markup
  style.css       the frame around the canvas
  js/engine.js    copied unchanged from Overdrive - the ninth game on it
  js/moles.js     the scheduler and the hit test - no pixels, no engine, no DOM
  js/burrow.js    the playable layer
  test-moles.js   node test-moles.js

It is a scheduler with a coat of paint

Everything the player experiences — whether it feels fair, whether it feels random, whether it gets harder in a way they can rise to — is decided by which hole pops next, when, and for how long. Write that as pick a hole at random, show it a bit less time each round and you get three separate faults, none of which look like bugs:

The same hole pops twice running. It reads as the game glitching rather than as luck, because the player's eye is already leaving it. The picker never returns the hole it just returned — 20,000 draws, zero repeats.

Uniform random produces droughts. A hole that has not been used for twenty pops is not a surprise, it is an annoyance. Holes are weighted by how long they have waited, so nothing is starved. Measured against a uniform picker on the same seed, the longest any hole sits idle drops from 30 pops to 21, and use is even to within 1.25×.

Shrinking the up-time eventually shows a mole for less time than a human can react to. That is not difficulty. It is a broken game that looks like a hard one.

$ node test-moles.js
43 passed, 0 failed

Difficulty by density, not by impossibility

Simple visual reaction time is about 250 ms — for someone expecting the signal, with no aiming involved. Add moving a finger to a target and 250 ms is the floor, not the average.

So up-time only tightens from 1.65s to 0.70s across the whole ramp, and the real difficulty knob is how many are up at once, which goes from one to five. The test asserts that relationship directly: density multiplies by more than speed divides by.

There is also a hard clamp at 0.62s that, as the tests confirm, never actually fires — the curve settles above it on its own. That is the point of it. If somebody later retunes the ramp and pushes it under a human's reaction time, the test fails and tells them, and the clamp means the game degrades safely rather than becoming unplayable while they work it out.

Two details that decide how it feels

A mole is drawn into its hole. Everything below the rim is clipped away and the rim is painted back over the top, so it rises out of the ground instead of fading in above it. The test checks a half-risen mole draws nothing in the band above the hole — without the clip, a mole halfway up looks like a mole floating over a hole, and the whole illusion goes.

A bomb that gets away is not a miss. Not hitting it is the correct play, so letting it go must not break the chain. Only a mole that escapes does that.

Each kind differs in shape as well as colour — the bomb has a lit fuse, the golden mole a crown — so nothing depends on telling brown from gold. Verified by reading the rendered pixels back off the canvas.

Restraint is worth two and a half times the score

Three bots over the same sixty seconds:

The indiscriminate bot hit five bombs and lost 25 of its 60 seconds to them. Knowing what not to hit is worth more than reaction speed.

Scoring

Missing a mole breaks the chain too. Best score is kept in localStorage.

Controls

Tap, or click. That is the whole game. R starts over.

Things deliberately left undone

• No sound, which this one misses more than most.
• The grid is fixed at 3×3. Config.COLS/ROWS and the picker both take any size.
• No mole that needs two hits, which is the usual way to add texture once density has run out of room.
• The run is always sixty seconds. An endless mode with three lives is the other obvious shape.