This is an earlier experiment. Read the amendments below before treating its findings as current.
The question
Which of the wing's seven agility differences from the ideal craft (instant response) drives its calm losses, and can wider spacing compensate?
What was tested
84 calm runs each reset one wing limit to the ideal value. 60 varied the response lag (the delay before it answers a command) from 0 to 1 s. 36, plus 12 ideal-craft checks, widened the lag allowance in the spacing from about 29 m to 53 m.
What was found then
Removing the lag cut losses from 532 to 56. Removing the vertical acceleration limits came second (−41 %); the sideways limit never comes into play, and removing the horizontal one made things worse (628). Losses rose with every quarter-second of lag, with no threshold, first read as about 40 craft per 0.1 s. Wider spacing helped less (best 320, widest 344) and cut completed laps 9–16 %. Read then as: distance does not buy back time.
What we learned later
Withdrawn: 'the steering plans as if the airframe answers instantly'; the spacing rule already allows for lag. The 90 % cut holds only in this setup, and changing lag also changed spacing and spread.
research/envelope.md:2757
Narrowed: 'rises at every step' does not mean in a straight line; the 40-per-0.1 s figure only describes these runs.
research/envelope.md:2824
Narrowed: 'levelling off' rested on the last setting alone, not a plateau; still wider spacing gave 290 and then 406 losses.
research/envelope.md:2879; research/envelope.md:2927
What changed
Nothing. A setting that widens the lag part of the spacing was added, off by default.
Numbers and source references
- seven
- agility settings where the wing differs from the ideal craftresearch/envelope.md:2757
- 84
- runs in the one-setting-at-a-time testresearch/envelope.md:2757
- 532 → 56
- wing craft lost per 12 runs with the lag removedresearch/envelope.md:2757
- 0.8 s / 0.2 s
- the wing's response smoothing and pure delayresearch/envelope.md:2757
- −41 %
- loss change without the wing's vertical acceleration limits (314 lost)research/envelope.md:2757
- 628
- losses without the wing's horizontal acceleration limitresearch/envelope.md:2757
- 60
- runs in the lag doseresearch/envelope.md:2824
- 56 / 160 / 264 / 414 / 532
- losses per 12 runs at total lag 0 / 0.25 / 0.5 / 0.75 / 1 sresearch/envelope.md:2824
- about 40 per 0.1 s
- the cycle's first reading of the lag costresearch/reviews/brief-c76.md
- 36 + 12
- wing runs at spacing coefficient 2 / 3 / 4, plus ideal-craft checksresearch/reviews/brief-c77.md
- 29 → 53 m
- horizontal separation from coefficient 1 to 4research/reviews/brief-c77.md
- 532 → 422 → 320 → 344
- losses per 12 runs at coefficient 1 / 2 / 3 / 4research/envelope.md:2879
- 9–16 %
- fall in completed laps with wider spacing (post hoc)research/envelope.md:2879