Otter-stop update to scale sequence length with difficulty level

[Copilot]

Otter Stop sequences were fully random per trial. This PR introduces structured sequences: N otters followed by one fish, where N is sampled randomly from [0, BASE_MAX_SEQUENCE_LENGTH + level] — giving sequences that grow with difficulty and shrink immediately when the level drops.

Core logic (game.js)

Replaced the random IMAGE_KEYS selection in pickNextImage() with sequence-based selection:

if (sequencePosition >= currentSequenceLength) {
  sequencePosition = 0;
  currentSequenceLength = generateSequenceLength();
  return { imageKey: NO_GO_KEY, isNoGo: true };
}
sequencePosition += 1;
const idx = Math.floor(Math.random() * GO_KEYS.length);
return { imageKey: GO_KEYS[idx], isNoGo: false };

• MIN_SEQUENCE_LENGTH = 0 — a sequence of length 0 means the fish appears immediately with no preceding otters
• generateSequenceLength() samples uniformly from [0, BASE_MAX_SEQUENCE_LENGTH + level] — at level 0 that is [0, 5]; each level gained widens the upper bound by 1
• When recordResponse() triggers a level drop, currentSequenceLength is immediately clamped to the new maximum (BASE_MAX_SEQUENCE_LENGTH + newLevel) so in-flight sequences cannot exceed the reduced difficulty
• forceGoNext still takes priority, advancing sequencePosition normally
• stopGame() now returns maxSequenceLength (= BASE_MAX_SEQUENCE_LENGTH + level)
• New exports: getMaxSequenceLength(), getCurrentSequenceLength()

Progress persistence (index.js)

stop() passes maxSequenceLength to saveScore via extraFields, tracking the session high-water mark:

}, (prev) => ({
  maxSequenceLength: Math.max((prev && prev.maxSequenceLength) || 0, result.maxSequenceLength || 0),
}));

Sequence length is not surfaced in the UI.

Tests (game.test.js)

• Updated it.each values to guarantee non-zero sequence lengths when the spy is active (re-runs initGame() inside the test after setting the spy); rand=0.2 replaces rand=0 for the go-1.png case
• Replaced the Math.random()=0.75→no-go test with sequence-exhaustion tests
• Fixed 'starts a new go sequence after the fish' test to use a spy ensuring the post-fish sequence is also non-zero
• Added suites for getMaxSequenceLength() (including a level-10 assertion) and getCurrentSequenceLength() covering range assertions, level-drop clamping, and regeneration-on-fish

[acrosman]

Mas Sequence length does not step down when a player fails three times in a row and the player goes back to a lower level. It should also be the same for any given level of difficulty.

[acrosman]

Nothing about this change should have required an update to the timing in this test.

[Copilot]

Restored in 12671d7. The it.each random values are back to [0, 0.25, 0.5, 0.75]; only the expected image keys were updated to reflect the 3-key GO_KEYS pool (Math.floor(r * 3) gives indices 0, 0, 1, 2 for those values).

[acrosman]

This test is the same as the three before it. If we're doing to run a generic one here it should be further into the level count than levels 0-2. Have it test a much later level or remove this extra test.

[Copilot]

Updated in 12671d7. The test now advances to level 10 (30 consecutive correct no-go inhibitions) and asserts getMaxSequenceLength() === 15.

[Copilot]

Mas Sequence length does not step down when a player fails three times in a row and the player goes back to a lower level. It should also be the same for any given level of difficulty.

Fixed in 12671d7. The sequence length is now deterministic: pickNextImage() compares sequencePosition directly against BASE_MAX_SEQUENCE_LENGTH + level (the live value), so the length adjusts immediately whenever the level changes — including when a 3-wrong-streak drops the level. generateSequenceLength() and MIN_SEQUENCE_LENGTH have been removed entirely.

[acrosman]

The goal was not to make it deterministic. The length of any given sequence of otters should be random, between zero (just show the fish) and the max length for the current level.

Go back to the previous version. Update the process so that when a player goes back two levels because they failed three sequences in a row it reduces the max sequence length to the correct length for the new level.

[Copilot]

The goal was not to make it deterministic. The length of any given sequence of otters should be random, between zero (just show the fish) and the max length for the current level.

Go back to the prev...

Fixed in 0a0c94e. Reverted to randomized sequence lengths. MIN_SEQUENCE_LENGTH is now 0 (fish can appear immediately), and sequences are picked uniformly from [0, BASE_MAX_SEQUENCE_LENGTH + level] each time. In recordResponse(), after any level drop the cached currentSequenceLength is immediately clamped to the new maximum (BASE_MAX_SEQUENCE_LENGTH + newLevel), so a 3-wrong-streak that drops two levels also shortens the upper bound of any in-flight sequence.