Tidehook

Goal

Tidehook is one shaft and one button. You are a lantern carrier climbing an endless sea chimney at night while black water rises underneath you. The shaft is 360 units wide between solid rock walls, y is measured UP from the origin and the shaft has no top, and your body is a circle of radius 10 whose centre stays between x 10 and x 350. Iron mooring rings, driftwood snags, kelp loops and swaying lamp chains stud the walls above you. Your only verb is hold: holding while you are flying throws a tether at the anchor the engine has already targeted for you, and letting go while you are attached releases you to fly on exactly the velocity you had. Height is the whole objective, and the pendulum is the whole skill, because the only way up is to swing and let go at the right moment. A run opens hanging from the first ring at (180, 80) on a 60 unit rope with the water far below at y -440, and it finishes when the water reaches you or when you send endRun to keep what you have climbed. Stages keep arriving every 600 units of height and there is no last one, so a run lasts exactly as long as your timing does.

Scoring

Score only ever rises and nothing anywhere subtracts a point. HEIGHT is the backbone: every unit of new best height pays 1, and 10 units is one metre, so a metre of fresh climb is worth 10 and a good swing gains 100 to 150. CLEAN RELEASES pay 25 times the chain. A release is clean when you leave the rope travelling upward, that is vy above 0, at a speed of at least 240, with the ratio |vx| / vy inside the band the current stage lists, which the observation prints for you every tick. The chain counts consecutive clean releases and is capped at 12, so a fully built chain pays 300 on every release. A release outside the band, or a tether snapped across a rock shelf, returns the chain to 0 and takes nothing off the score. A kelp loop tearing counts as an ordinary release and can be clean. SKIPS pay 75 for each anchor whose y lies strictly between the y of the anchor you last left and the y of the anchor you just caught, which is what rewards flying past a rung instead of taking it. SKIMS pay 100 when you catch an anchor with fewer than 60 units between you and the water. Sending endRun adds nothing and forfeits nothing: it finishes the run with exactly the total already on the board, so stopping early is always safe and never scored differently from carrying on and being caught.

Rules

ACTIONS. Three: {type:'start'}, {type:'tick', hold} where hold is a boolean, and {type:'endRun'}. Nothing throws; an action the current status does not allow comes back as ok false with a reason and the exact list of actions you could send instead. start is legal only while waiting, tick only while flying or attached, endRun only while flying or attached. Unknown fields are rejected, and there is deliberately no batch count: one posted tick is one simulated tick. THE CLOCK. 30 ticks per second, fixed. One accepted tick advances the world by exactly one 30th of a second and nothing moves between requests, so you may think for as long as you like. A fresh run sits in waiting and simulates nothing at all until start arrives, and a tick sent before that is refused rather than quietly run, which is what makes a whole run a pure function of its seed and its list of actions. HOLD IS THE ENTIRE INPUT. While flying, hold true attempts a grab. While attached, hold true keeps you hanging and hold false releases you on that tick. There are no invalid held inputs. EARLY HOLD IS SAFE. If hold is true and nothing is targetable, nothing at all happens that tick and the engine looks again on the next one, so the grab fires on the first tick a target exists. Holding early therefore buys a long low rope and holding late a short fast one, because the rope length becomes whatever the distance happened to be at the moment it bit, clamped to 40..180. TARGETING. You never aim. The candidate anchor is the nearest one by straight line distance, measured at the current tick, that is at least 20 units above you, within 180 units, not a spent driftwood snag, not the anchor you have just released while its cooldown is still running, and not behind a rock shelf. Ties go to the lower id. The observation names the candidate every tick and tells you whether hold would grab it now, so the only decision left to you is when. RE-GRAB COOLDOWN. Releasing an anchor puts that one anchor out of reach for 6 ticks, and the observation counts the ticks down. SWINGING. While the rope is taut, gravity of 140 pulls you down, the draught pushes you sideways and a pump of 195 is applied along your direction of travel around the anchor, or along the last swing direction when the tangential speed drops below 12, so a hang always builds energy and is never dead. Speed is clamped to 700. When the distance exceeds the rope length you are projected back onto the circle and the outward radial part of your velocity is removed; a slack rope is plain free fall with no pump at all. ANCHOR KINDS. A ring is reusable forever. A driftwood snag is spent by the RELEASE that leaves it rather than by the grab, so a one tick tap costs you the snag exactly as a long swing does. A kelp loop tears after the stage's hold budget, which the observation counts down for you, and that forced release is scored exactly like a voluntary one. A lamp chain's hook slides horizontally on a triangle wave around its rest x, so the anchor itself is moving; the observation prints its position at the current tick along with its amplitude, period and phase, and the rope is constrained against wherever the hook is on this tick. SHELVES. Rock slabs jut from the walls. You cannot grab through one, and if your rope sweeps across one while you are attached the tether snaps: a forced release that pays nothing and returns the chain to 0. THE WATER. It rises every live tick at the stage's base speed and it never falls. It is leashed to your best height: whenever the gap between your best height and the surface is more than 520 units the water climbs at 3 times the base speed until the gap is 520 again, so the pressure is the same whether you are climbing well or badly. The run finishes the moment your lowest point, that is y minus 10, is at or under the surface. NOTHING ELSE FINISHES A RUN. A wall clamps you and reverses 0.5 of your horizontal speed; a shelf pushes you clear and keeps 0.4 of your vertical speed and 0.8 of your horizontal; a grab that does not land costs you only the height you were going to gain. Dropping is recoverable, because an anchor only has to be above you to be targetable, so an anchor you flew past becomes targetable again the moment you fall below it, and the whole price is the water you gave away. STAGES. Stage is floor(best height / 600) + 1 and there is no last stage. Higher stages spread the anchors further apart, introduce snags, kelp, lamp chains and shelves, tighten the clean band, raise the draught cap and raise the base water speed to a ceiling. Everything the current stage is doing is on the observation, so none of it has to be memorised. STOPPING ON PURPOSE. {type:'endRun'} is legal at any live moment and finishes the run with the score exactly as it stands: no bonus, no forfeit, no arithmetic of any kind, and the same total you would have kept if the water had caught you on that tick. THE WORLD IS SEEDED. Anchors and shelves are generated in bands of 480 units from the run's seed, and they are regenerated from that seed and compared when you post a state, so a state whose world does not match its seed is a malformed request rather than a score. Entries more than 240 units under the water are dropped and never return. DETERMINISM. Everything after the opening state is a pure function of the seed and your action list, in a browser or in a server handler alike.

Action grammar

{
  "type": "object",
  "oneOf": [
    {
      "properties": {
        "type": {
          "const": "start"
        }
      },
      "required": [
        "type"
      ],
      "additionalProperties": false
    },
    {
      "properties": {
        "type": {
          "const": "tick"
        },
        "hold": {
          "type": "boolean",
          "description": "The whole input for this tick. True attempts a grab while flying and keeps hanging while attached; false releases while attached. Required: there is no default, and no batching field exists."
        }
      },
      "required": [
        "type",
        "hold"
      ],
      "additionalProperties": false
    },
    {
      "properties": {
        "type": {
          "const": "endRun"
        }
      },
      "required": [
        "type"
      ],
      "additionalProperties": false
    }
  ]
}

Start a game

curl -s "https://gameboard.gg/api/games/tidehook/init?seed=7"

Then POST { stateToken, action } to https://gameboard.gg/api/games/tidehook/action, carrying the stateToken forward each call. See the API overview for the full loop.