## Executed airborne-flip supplement (B_air)

**Breakout has the highest nominal speed found in this bounded airborne-dodge search plus equal-budget cross-preset transfers: 132.448 kph.** This is not a ground-start jump/flip result, a rigorous optimum, or Rocket League validation. Plank and Hybrid are both included; Plank is the existing 2016 Batmobile preset.

| Preset | Independent search (kph) | Best incl. transfers (kph) | Winning schedule source | Δ Octane (kph) | Δ Octane (%) | +X component (kph) | Yaw / elevation ° | Near-forward / diagonal peaks (kph) | Surface | Episodes / callbacks | Valid execution errors |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Octane | 127.869 | 128.301 | Breakout | +0.000 | +0.000 | 128.025 | 3.58 / 1.14 | 127.869 / 128.301 | front-face | 1 / 2 | 92.9% |
| Dominus | 125.603 | 131.492 | Plank | +3.191 | +2.487 | 131.414 | -0.40 / 1.93 | 131.492 / 128.985 | front-edge | 1 / 2 | 85.7% |
| Plank | 132.330 | 132.330 | Plank | +4.029 | +3.140 | 132.269 | -0.29 / 1.71 | 132.330 / 129.194 | front-edge | 1 / 2 | 85.7% |
| Breakout | 129.209 | 132.448 | Plank | +4.147 | +3.232 | 132.381 | -0.29 / 1.79 | 132.448 / 129.209 | front-edge | 1 / 2 | 85.7% |
| Hybrid | 123.293 | 128.425 | Breakout | +0.124 | +0.096 | 128.164 | 3.49 / 1.09 | 128.187 / 128.425 | front-face | 1 / 2 | 92.9% |

### How the flip is legal without arena meshes

Use `THE_VOID` with unmodified Soccar car/ball physics. Start in a nonpenetrating airborne state with an unused air dodge: `is_on_ground=False`, `has_jumped=False`, `has_double_jumped=False`, `has_flipped=False`, zero angular velocity, initial car velocity (1800,0,0) uu/s (**64.8 kph**) and 100 boost. This is an airborne state of the kind available after leaving a surface without jumping or obtaining a flip reset; its prehistory is not simulated or validated in Rocket League. No wheel-ground contact, fake platform, first-jump impulse, or angular velocity is injected. Winner audits also verify no suspension-wheel contacts through first impact and no post-tick car speed exceeding the stock cap. The jump button is pressed for one tick with directional pitch/yaw, so RocketSim itself creates the dodge impulse and rotation; earlier pitch/yaw/roll controls can orient/spin the car legally.

The ball is fixed initially at (0,0,1100) uu with v=(0,0,−0.001) uu/s and zero spin. This tiny nonzero initial falling speed prevents the known stationary-ball sleeping shortcut; gravity remains −650 uu/s². Every preset shares the same ball, initial car velocity and boost. Geometric alignment uses the rotated front-face-center tangent, followed by identical world-axis search displacements.

Twelve bounded parameters: additional X gap50–800 uu, lateral±80 uu, ball-above-hitbox-center displacement±150 uu, initial yaw±15°/pitch±30°/roll0°, dodge tick0–50, direction±60° (0=forward), pre-dodge pitch±1 and yaw/roll±0.5, boost duration0–120 ticks, post-dodge pitch/cancel±1. Throttle1; steering/handbrake0 in air. Window≤240 ticks (2 s), ended earlier if either body leaves conservative central airspace. This boundary is an observation limit, not a physical wall. First contact must occur after the dodge input has taken effect; exactly one completed contact episode, horizontal±5° and elevation0–30° required. All input schedules and exact initial states are saved.

30,000 independent search candidates = 5 presets ×6000; seed20260909, 243 coarse candidates plus the same DE optimizer as A, 212.58s search runtime. Then 25 transfer evaluations (five original winning schedules per target, including its own replay) form a second ranking stage. Promote only valid improvements, save them separately in `B_air_transfer_*/winner.json`, replay exactly, and run sensitivity around the promoted states. Original independent search winners remain unchanged; `air_transfer_selection.json` preserves all transfer proposals. The coarse grid includes 0° and ±45° dodges equally. Reported direction buckets are descriptive: near-forward |direction|≤10°, diagonal >10°; these are not independently optimized categories. Every winner is replayed, receives ±1 tick/±2 uu/±0.5° errors, and transfers to every other preset with only documented geometric realignment. `air_results.csv`, `air_measurements.csv.gz`, per-condition JSON and `cross_preset.json` preserve measurements.

### Pivot motion and important speed-clamp caveat

| Preset | Incoming car v (kph) | Incoming ball v (kph) | ω world (rad/s) | Contact r world (uu) | ω × r · n (kph) | Contact v · n (kph) | Extra Δv world (kph) | Dodge / first contact tick | Transient car speed >82.8 kph |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Octane | (95.99, -8.80, -6.38) | (0.00, 0.00, -7.57) | (1.14, -0.40, 0.82) | (72.34, 2.88, 18.48) | -0.009 | 92.237 | (48.95, 3.00, 1.35) | 39 / 39 | YES |
| Dominus | (98.21, 7.40, -8.95) | (0.00, 0.00, -9.30) | (1.10, 0.95, 0.33) | (77.47, -2.24, 21.85) | 0.554 | 97.867 | (48.86, -2.39, 7.58) | 48 / 48 | YES |
| Plank | (98.21, 7.40, -8.95) | (0.00, 0.00, -9.30) | (1.10, 0.95, 0.33) | (77.38, -1.95, 17.20) | 0.360 | 97.422 | (48.95, -2.40, 6.96) | 48 / 48 | YES |
| Breakout | (98.21, 7.40, -8.95) | (0.00, 0.00, -9.30) | (1.10, 0.95, 0.33) | (82.18, -1.76, 16.70) | 0.302 | 97.194 | (48.97, -2.39, 6.86) | 48 / 48 | YES |
| Hybrid | (95.99, -8.80, -6.38) | (0.00, 0.00, -7.57) | (1.14, -0.40, 0.82) | (76.89, 2.86, 18.79) | 0.011 | 92.257 | (48.97, 2.82, 1.24) | 39 / 39 | YES |

Vectors use world axes, except where explicitly marked; velocity display is kph, raw JSON calculations use uu/s. Rotation before impact comes from controls, not state injection. The angular term is measured as `(omega × contact_offset) · normal`; total contact speed also includes pivot velocity. Contact locations/normals are geometry-reconstructed and cross-checked against the actual sphere-side callback, as in A.

**Stock RocketSim may temporarily exceed the 2300 uu/s (82.8 kph) car cap inside a dodge tick.** `_UpdateDoubleJumpOrFlip` applies its immediate impulse before collision solving; `_FinishPhysicsTick` clamps speed afterwards. The table exposes above-cap incoming callback speeds and dodge/contact timing. In this run, every selected winner hits on the dodge-input tick. `_UpdateAirTorque` runs before the new dodge is registered, so the first-contact angular velocity was built by earlier air controls, not by a completed flip rotation. These are **legal dodge-assisted hits**, not proof that the rotational phase of a flip supplies the observed gain. A separately optimized study requiring contact at least two ticks after the dodge has not been executed. This makes ±1-tick timing sensitivity particularly important. Such shots are legal in this simulator but depend on this within-tick ordering; they are not evidence the same transient collision exists identically in Rocket League. The cap was not raised or disabled. Ball cap remains216 kph. Approach-speed differences, contact orientation, extra car-ball impulses and inertia prevent attributing every speed difference to hitbox reach or rotation alone.

**Transfer results undermine a simplistic hitbox ranking:** the original Plank search sequence also produces a faster valid shot on Breakout than Breakout's independent optimizer found. Several other presets improve using another preset's sequence. The final table includes these measured improvements; it does not claim an intrinsic Plank/Breakout advantage or that the remaining search gap is irreducible.

A's winning no-flip conditions and B_air have different initial ball/car velocities and setup bounds. Compare presets **within** each category; do not interpret A-to-B_air gains as a clean causal effect of flipping.

### Execution sensitivity and no-dodge control

| Preset | Nominal (kph) | Diagnostic peak, all (kph) | Diagnostic peak, valid (kph) | Median (kph) | Minimum (kph) | Maximum (kph) | Valid fraction |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Octane | 128.301 | 128.610 | 128.610 | 128.286 | 107.915 | 128.610 | 0.929 |
| Dominus | 131.492 | 131.578 | 131.578 | 131.457 | 111.737 | 131.578 | 0.857 |
| Plank | 132.330 | 132.396 | 132.396 | 132.309 | 112.547 | 132.396 | 0.857 |
| Breakout | 132.448 | 132.554 | 132.554 | 132.434 | 112.679 | 132.554 | 0.857 |
| Hybrid | 128.425 | 128.696 | 128.696 | 128.412 | 108.020 | 128.696 | 0.929 |

Median/range use nominal plus all14 errors; valid fraction excludes nominal. Some errors miss the same-tick dodge/contact coincidence or violate the cone. A shift earlier than tick0 cannot reproduce a dodge before the defined initial state; it is outside the available input window. Sensitivity is not simulator-accuracy uncertainty. Higher valid diagnostic speeds are published separately and not silently substituted for optimization-stage winners.

| Preset | With dodge (kph) | Without dodge, unoptimized (kph) | Without dodge in forward cone |
| --- | --- | --- | --- |
| Octane | 128.301 | 108.319 | True |
| Dominus | 131.492 | 112.328 | True |
| Plank | 132.330 | 113.032 | True |
| Breakout | 132.448 | 113.134 | True |
| Hybrid | 128.425 | 108.417 | True |

This removes only the jump input from each exact winning sequence, keeping its initial state and other controls. It is an **unoptimized control ablation**, not a fair best-no-flip ranking; the no-dodge ball can leave the forward cone and is not a valid B_air candidate.

### Running it, or switching to ground-start flips

`.venv/bin/python benchmark.py run --category B_air --budget 6000` works now without meshes. Follow with `.venv/bin/python rank_air_transfers.py`, then `audit_replays.py` and `visualize.py` to reproduce the transfer-pooled ranking and visuals. For ground-start jump/dodge timing, dump genuine Soccar meshes on Windows with RLArenaCollisionDumper and run `--category B --meshes collision-meshes --budget 6000`. Grounded B and carry/flick C remain unexecuted; no airborne result is labeled as either.