AutoMovie API
    Preparing search index...

    Namespace HumanSkill

    A performer's skill as numbers laws read: each proficiency axis resolved to [0, 1], the interpolation every family's quantities use, the seeded draw streams, and the few variability laws that hold across tasks.

    A quantity is lerp(novice, elite, smoothstep(axis)), so skill eases in and out at the ends of the scale. The novice and elite values belong to the action family whose studies measured them (striking speeds and phase durations to the strikes, gait speed and stride variability to locomotion, movement time and scatter to reaching and grasping, reaction times to defence and orienting); a family never inherits another family's numbers, and a family without a located study states provisional values for capture review. Only these hold across tasks and live here:

    • timing variability, Wing and Kristofferson's (1973) two-level model sigma^2 = motor^2 + (CV T)^2 for any timed interval, with a trained CV about half an untrained one (Grondin and Killeen 2009): CV 7 % and motor SD 25 ms to CV 2.5 % and 10 ms on the timing axis;
    • endpoint scatter growing with distance over time, impulse variability (Schmidt et al. 1979; Harris and Wolpert 1998), sigma = k D / T with k from 0.012 s to 0.004 s on the path axis (provisional constants);
    • postural sway RMS, a factor of 1.6 to 1.0 on the balance axis (experts sway less, Perrin et al. 2002; factor provisional).

    Draws come from counter-based streams (createAutoMovieRandom, Philox, Salmon et al. 2011) keyed by the chain seed, the performer, the action, its instance and the channel, so adding or removing one action never changes another's draws. Skill never raises a range of motion.

    AXES
    LEVELS
    jitter
    normal
    quantity
    reaction
    resolve
    scatter
    series
    stream
    sway