Task 3 (Floater)

Wind estimated from every pilot’s circling (centre drift preferred, ground-speed modulation as fallback), vector-averaged two ways: by hour of day (how the wind built and shifted through the day) and by speed-section leg (the wind each part of the course saw). Field-level only — no per-pilot value.

All pilots’ thermal climbs bucketed by the hour the climb started (labelled in the competition’s time zone): median and 90th-percentile average climb rate per hour show how the day’s lift developed. Field-level only — no per-pilot value.

Share of a pilot’s airborne time (on the shared grid) spent in non-sinking air — 30 s-smoothed vario at or above −0.5 m/s. Reflects how much of the flight met usable air: when flown, where the pilot steered, and how the flight ended all feed it, so the correlation sign is the finding. The timing table relates the day’s best climbs window to when the field actually launched.

In every thermal two or more pilots used, each use is ranked by average climb rate; a use’s percentile is the share of strictly slower uses. The value is the duration-weighted mean percentile over the pilot’s shared climbs — centering skill isolated from thermal selection.

For each thermal of at least 60 s, the seconds from entering until the 30 s rolling climb rate first reaches 90% of its peak in that thermal. The value is the median across the pilot’s thermals — how fast they centre the best lift after arriving.

For each thermal of at least 90 s, the climb rate over its final 30 s — the "give-up rate". The value is the median: low means the pilot abandons weakening lift early, high means they ride climbs to the end; the correlation sign says which behaviour paid on this task.

Post-start circling bouts of at least 30 s are lift encounters; one that overlaps a detected thermal was accepted, otherwise the pilot circled and rejected the lift. The value is the percentage accepted — low means picky, high means they take almost everything they turn in.

The altitude at which the pilot left each post-start thermal, as a percentage of the day’s working band (0% = field floor, 100% = field ceiling); the value is the median. High means topping out every climb, low means leaving early with climb still available.

Each detected circle is least-squares fitted; its RMS fit error divided by the fitted radius measures how round the turn really was. The value is the median over all the pilot’s circles — lower means smoother, more consistent turning.

Duration-weighted mean ground speed over post-start glide segments (glide distance ÷ glide time). Higher means the pilot covers ground faster between climbs.

For each completed speed-section leg, the pilot's glide-phase L/D (path distance ÷ net altitude lost while gliding; legs losing under 100 m skipped) is divided by the field's median L/D on that same leg, then averaged over legs. Above 1.0 means the pilot found better glide lines than the field on the same legs.

A speed-to-fly PROXY — no glider polar data exists. Each post-start glide is paired with the climb rate of the next thermal starting within 5 minutes; the value is the mean glide speed flown before stronger-than-median climbs minus the mean before weaker ones. Positive means the pilot speeds up when the next climb justifies it.

Route distance flown on each completed speed-section leg — full path outside climbs, each climb counted as its net drift — ÷ the leg's optimized distance, averaged with optimized-distance weights. Closer to 1.0 means less deviation from the optimal course line; circling path is excluded, so stopping to climb does not read as bad line choice.

Share of post-start altitude gain (10 s-smoothed) made outside detected thermals — climbing on the run instead of stopping to circle. Neutral: the correlation sign shows whether dolphin climbing paid on this day.

Finds each pilot's post-start altitude minima (30 s-smoothed, at least 100 m below the surrounding maxima) — the points where they stopped descending and committed to a climb — and reports the median as a percentage of the day's working band. Higher means the pilot keeps more margin in hand before taking a climb.

Counts post-start climbs entered below 15% of the working band that went on to gain at least 300 m — genuine low saves. Zero is a real score for a started pilot; the correlation sign says whether digging out or never getting low is what pays.

Post-start thermal count per 100 km of scored flown distance — how often the pilot stops to circle. Each pilot's note adds their mean climb percentile within shared thermals, so few stops can be read together with climb strength.

Share of speed-section time (start to ESS, or landing) spent searching — neither climbing in a thermal nor gliding with real net speed. Lower means less time leaks away between climbs.

Share of a pilot's post-start flying time spent inside a detected gaggle (clustered with at least one other racing pilot on the shared time grid). Neutral: the correlation sign says whether sticking with company paid on the day.

Share of a pilot's post-start climbs where another pilot was already established (at least 30 s) and still climbing in the same thermal when they joined. High = climbs on others' markers; low = finds their own lift.

When a pilot leaves a gaggle that keeps flying, did leaving pay off? We compare the leaver's arrival at the next turnpoint against the median arrival of the pilots who stayed. Win rate > 50% means their departures beat the gaggle. Only pilots still in the gaggle after the split who reached that turnpoint after it count as stayers.

Seconds from the start gate taken (or the pilot’s own crossing on elapsed-time tasks, where the delay is 0 by definition) to the scored SSS crossing. The start table adds crossing altitude and how far behind the leading already-started pilot each start was.

For each completed speed-section leg, the pilot’s leg time is compared with the mean of the top-10 (by rank) pilots who completed that leg; only losses count, and the losses are summed. Summed leg times are race time and the reference is defined by rank, so this tracks the outcome by construction — read the waterfall table (every leg against the task winner) for the diagnostic, not the correlation for a finding.

At each speed-section turnpoint, elapsed race time (reaching minus own start) is compared with the fastest pilot to that turnpoint; the scalar is minutes behind at ESS. Expected to track final rank almost perfectly — this metric is the eval’s sanity check.

Altitude at ESS minus the altitude needed to glide to goal at the sport’s standard glide ratio (S7F §12.3.6: 5.0 HG / 4.0 PG). A big positive margin is altitude — i.e. time — left unspent.

At the pilot’s last climb before ESS (or landing): distance to goal divided by height above goal — the glide ratio they committed to. Only counted when that climb ended within 1.5× the final leg’s distance of goal.