diff options
Diffstat (limited to '')
| -rw-r--r-- | platformer/movement_feedback_1.md | 142 |
1 files changed, 142 insertions, 0 deletions
diff --git a/platformer/movement_feedback_1.md b/platformer/movement_feedback_1.md new file mode 100644 index 0000000..672218b --- /dev/null +++ b/platformer/movement_feedback_1.md @@ -0,0 +1,142 @@ +# Movement feedback, round 1 + +The first round of playtest feedback on movement, and the changes we made in response. The numbers +here are as of this round. `src/config.h` is the source of truth, and `game_feel.md` holds the full +feel guide and tuning history. + +--- + +## What players reported + +1. **The second jump's timing was too hard.** Players pressed A a hair early, while still falling, + and the press was lost: no jump happened on landing, so the second-jump window slipped by. +2. **Players expected to wall hang on a single jump.** Only the second jump, retreat jump and wall + jumps could grab walls, so running at a wall and jumping once did nothing. + +Two smaller requests came with them: + +3. Stopping should be very slightly quicker when the stick is released. The skid stays as it is. +4. Starting a skid (`retreating_movement`) should need a slightly more deliberate pull-back. + +--- + +## What changed + +### 1. A 0.1 s jump buffer (the second-jump timing) + +An A press is now remembered for `PLAYER_JUMP_BUFFER` (0.1 s). The first moment a jump is possible +within that time, it happens. + +- **On the ground** it's instant, just as before. +- **Pressed while falling**, it fires on the landing tick. A press up to 6 ticks (0.10 s) before + landing still gets the second jump; 7 or more ticks early still does nothing. +- **Pressed just before grabbing a wall**, it fires an early wall jump the moment you grab. +- Holding A still decides hop vs. full jump. A buffered press that was already released gives a hop. + +Implementation: `Player.jump_buffer`, set in `player_update` on `jump_pressed` and counted down each +tick. Jumps start through `take_jump()`, which spends the buffer, in the ground, skid and wall-hang +states. + +### 2. Every jump can wall hang + +`heightened_movement` now turns on when **any** jump starts, including a plain first jump, and lasts +the whole jump. Landing still opens the usual 0.2 s second-jump window. A single jump into a violet +wall now grabs it. + +**A knock-on fix: the clear-the-top rule.** Once every jump could grab, hopping onto a ledge from just +below its lip would grab the wall face instead of landing on top. That's especially bad on the +hangable cliff. So now you don't grab a wall whose top your jump will carry you over anyway +(`would_clear_top()`: apex = y + vy² / 2g, compared with the wall's top). + +- A normal jump at the cliff from 1.2 m under its lip → lands on top. +- The same jump 10 m below the lip → grabs the face. + +This also fixed an older snag, where a second jump from just under the cliff lip grabbed the face. +To go back to strict "always grab", remove the one condition in `wants_wall_hang()`. + +### 3. A slightly quicker stop + +`PLAYER_GROUND_FRICTION`: 11.3 → **12.0** m/s² (about 6%). + +| From full speed (6 m/s) | Before | After | +|---|---|---| +| Stop time | 32 ticks (0.53 s) | 31 ticks (0.52 s) | +| Stop distance | 1.54 m | 1.45 m | + +The skid (`retreating_movement`) is unchanged. + +### 4. A more deliberate skid + +`PLAYER_SKID_ANGLE_DEG`: 120° → **130°**. You have to pull the stick more squarely back to skid. + +| Pull-back from the run direction | 120° | 125° | 130° | 135° | 140° | +|---|---|---|---|---|---| +| Result | run | run | run | **skid** | **skid** | + +--- + +## What didn't change + +Measured and identical to before: the jump heights (2.33 m / 4.08 m second / 4.67 m retreat), the +0.2 s second-jump window, the skid (0.5 s glide-then-plant), the retreat jump, wall hang timing +(0.5 s, early wall jump in the first 0.1 s), mirrored wall jumps, the one-wall lockout, chimney +chaining, and riding moving platforms. The cliff course is still completable. + +--- + +## Things to watch in the next playtest + +- **Wall grabs are now everywhere.** Any jump into a violet wall or hangable rock grabs it. Near the + cliff and pillars that could feel sticky, or it could feel great. Watch for players getting + caught on walls they meant to jump past. +- **The clear-the-top threshold** is exact (apex vs. wall top). If players still snag on lips they + *nearly* clear, a small margin (e.g. grab only if the apex is at least 0.2 m below the top) would + help. +- **The buffer length.** 0.1 s fixes early presses in tests. If second jumps still feel strict, + 0.12–0.15 s is the next step. Going much longer risks jumps firing when players didn't mean them. +- **The skid angle.** 130° is stricter. If skids stop happening when players want them, back off + toward 125°. + +--- + +## Follow-up after playing it + +### Reverted: the first jump no longer wall hangs + +Change 2 went too far: **wall hangs off a single jump made them too strong.** A plain first jump no +longer turns on `heightened_movement` (it's back to the second, retreat and wall jumps, and wall +hangs). A first jump into the main wall no longer grabs it. + +**Kept:** the jump buffer (1), the quicker stop (3), the 130° skid angle (4), and the +**clear-the-top rule**. The rule was added for change 2, but it still earns its place: a second jump +from just under a ledge's lip lands on top instead of snagging on the face. + +### Cliff course regenerated + +Feedback: the platforms were too small (skidding off them), too crowded against the cliff, too +many needed double jumps (sometimes two in a row), and it wasn't clear which gaps needed a double. + +| | Before | After | +|---|---|---| +| Platform size | 3–4.5 m | **4–6 m** (all 39 measured in range) | +| Distance from the face | hugging it (0–9 m) | **1.5–13 m** (climb average 6.2 m); only the last few come in, for the hop onto the top | +| Heights | any value | **on a 1.5 m tier grid** from the cliff top: one tier up = a single jump, two = a double | +| Route steps (current seed) | mixed rises, doubles common | **19 singles, 8 doubles, 1 level; 0 back-to-back doubles** | +| Extra platforms | random spots and heights | each **one single hop beside a route platform**, level or one tier up or down | +| Rock pillars | placed first (and blocked the route) | placed after the route, never blocking one of its hops, 4–6 m out | +| Far cliff (the finish) | 60 m | **58 m**, two tiers above the top, so it's on the grid | +| Platforms in total | 37 | 39 (7 moving) | + +To fit a mostly-single-jump climb onto the cliff, the route **switches rows** (out a row, or back +in, at the same spot along the face) more often when it's getting too far along for its height. +The course test confirms that the top and the finish are reachable, that the bot lands every +still-platform hop it tried (30 of 30), and that all 7 moving platforms can be ridden. + +--- + +## Files touched + +`src/config.h` (new `PLAYER_JUMP_BUFFER`, friction, skid angle), `src/player.c` / `src/player.h` +(buffer, first-jump heightened and its revert, clear-the-top rule), `src/level.c` (the regenerated +cliff course: tiers, bigger platforms, row switches, pillars after the route, far cliff at 58 m), +`README.md` (controls table), and `game_feel.md` (rules, cliff course, tuning history). |
