Skip to main content

Task 3 (Open)

How the field flew this task, and which behaviours separated it.

View task
ELLIOTWALWALIGHTHKHANCO
The optimised route. Pilots fly it in the direction of the arrows. The radii, the leg distances and the start times are on the task page.

Analysis computed

Pilots
34
Airtime
85h (12:46–18:03 AEDT)
Thermals
24661 shared by 2+ pilots
Working band
866–2152 m
Airtime split
  • 43%climbing
  • 25%gliding
  • 32%searching

1 pilot is in the standings but not in this analysis. Which, and why

What the weather did

From the weather model

Independent of the tracklogs: modelled conditions for the task area.

Fetching the day’s weather — it will appear here in a moment.

From the pilots' tracks

What the field actually flew — wind, climb strength and leg timing measured from every pilot's tracklog.

The day’s wind, hour by hour and leg by leg. What the air did, read from the field itself. We estimate the wind from the circling of every pilot. The first method is the drift of the circle centre, and the second method, used when the first is not available, is the modulation of the ground speed. We then average the vectors two ways. The table by hour of day shows how the wind increased and changed direction through the day. The table by speed-section leg shows the wind on each part of the course. This metric describes the day, so it has no value for each pilot.

How strong the day’s climbs were, hour by hour. When the day started, reached its peak, and ended. We group the thermal climbs of all pilots by the hour in which each climb started, labelled in the time zone of the competition. The median and the 90th-percentile average climb rate for each hour show how the lift developed. This metric describes the day, so it has no value for each pilot.

Share of the flight spent in air that wasn’t sinking. How much of the flight was in air worth being in. The value is the share of the airborne time of a pilot, on the shared grid, with a 30 s-smoothed vario at or above −0.5 m/s. The time they flew, the line they steered and the way the flight ended all feed this value. It is therefore a reading of the day as much as of the pilot. There is no expected direction, and the sign of the correlation is the finding. The timing table compares the window of the day’s best climbs against the time when the field launched.

All charts — measured and modelled alike — share that one time axis, so a vertical scan compares the two at the same moment. Arrows fly WITH the wind — direction figures are degrees the wind blows from; arrow length and opacity track speed and sample count. On the per-leg chart the pale bar is when the field flew that leg and the solid band inside it is the circling its wind was measured from — a leg the field glided is measured in a sliver of the time it was flown. Exact numbers are in the day family’s tables under “The metrics in detail”.

The day's thermals

The 40 most-shared of 60 multi-pilot thermals, reconstructed by pooling every pilot's track through the same climb. Everything shown is measured from the tracks — no fitted lift model.

StartPilotsHeight bandMean climbStrongest sideDetail
5800–1400 m+1.3 m/sSW
41000–1700 m+1.0 m/sSE
4900–1200 m+1.1 m/sN
71200–2000 m+1.4 m/sSW
7900–1700 m+1.7 m/sNE
10800–1300 m+1.2 m/sNW
61600–1900 m+1.3 m/sNE
71100–2100 m+1.4 m/sS
6700–1400 m+0.8 m/sSE
51200–1600 m+1.2 m/sNW
10700–1900 m+1.5 m/sE
61700–2100 m+1.2 m/sSE
6900–1900 m+1.4 m/sSW
4800–1600 m+1.4 m/sNW
51300–1700 m+1.4 m/sSW
41500–2100 m+1.7 m/sE
41400–2200 m+1.6 m/sS
31200–1800 m+1.3 m/sE
31600–1900 m+1.2 m/sSE
51500–2000 m+1.1 m/sW
41200–2500 m+2.2 m/sSW
31300–1600 m+1.0 m/sNE
41700–2100 m+1.0 m/sSE
31800–2500 m+1.4 m/sW
51600–2400 m+1.5 m/sSW
31000–1700 m+1.9 m/sS
41700–2100 m+1.2 m/sSE
51700–2100 m+1.6 m/sSW
31500–2400 m+1.4 m/sNE
51200–2400 m+1.3 m/sNE
41000–2200 m+1.8 m/sS
32000–2500 m+1.4 m/sW
31100–1900 m+1.5 m/sNE
41600–2100 m+1.6 m/sN
31100–2300 m+2.4 m/sSW
42200–2500 m+1.5 m/sNW
31800–2300 m+1.8 m/sW
62000–2500 m+1.1 m/sN
31300–1800 m+1.0 m/sE
31300–1900 m+1.3 m/sSE

Thermal at 13:38 AEDT — 10 pilots, 22 climbs

  • Wind 10.3 km/h from 298° (WNW), measured from 96 circle estimates in the pilots' own tracks.
  • Model wind cross-check loading…
  • Leans 48° from vertical toward 90° (E), within 28° of downwind.
  • Strongest on the NW side of the core at +1.5 m/s against +1.2 m/s on the SE side.
  • Multiple cores in 2 of 5 bands between 900 and 1300 m — separate feeders (⬧ in the rose) before they merged.
Watch this thermal in the 3D replay (opens in a new tab)
Pilots in this thermal (climb rates)
PilotMinMedianMax
Adrian Connor-2.0 m/s+1.5 m/s+6.0 m/s
Howard Taylor-2.3 m/s+1.3 m/s+3.3 m/s
Richard Martin-2.3 m/s+1.3 m/s+3.5 m/s
Mick Lamb+0.5 m/s+0.5 m/s+2.0 m/s
Rob Larkin-1.0 m/s+0.5 m/s+1.5 m/s
Josh Woods———
Steven Crosby———
Rob de Groot———
Col Jackson———
Bradley Elliott-1.0 m/s-1.0 m/s-1.0 m/s

Each pilot's slowest, typical and best climb over their own vario samples in this thermal — a negative minimum means they touched sink inside it.

Band table (exact numbers)
BandCore offset E/N (m)Working radiusExtentMean climbBest climbSamplesPilotsCores
1200–1300 m177 / 23110 m151 m+1.5 m/s+5.0 m/s9322
1100–1200 m141 / 5461 m106 m+1.6 m/s+4.5 m/s11141
1000–1100 m76 / 27124 m189 m+1.4 m/s+4.0 m/s19431
900–1000 m-103 / 24148 m194 m+1.0 m/s+6.0 m/s25962
800–900 m-267 / 46145 m216 m+0.8 m/s+3.0 m/s6531

How to read this: each thermal pools every pilot's fixes through the same climb into 100 m altitude bands; a band's core is the lift-weighted centre of its fixes, so the rose and the sector readings are already normalised for the thermal's lean and drift. Wedge length is relative climb by side of the core; the dashed ring is the measured working radius and the dotted ring the widest the field ranged. The solid arrow is the wind measured from the pilots' circles; the dashed arrow is the weather model's wind for the same place, time and altitudes — a model run, not an observation.

Which behaviours went with better results

Every row is one behaviour, measured for each pilot and then compared against the published placings (Spearman's rank correlation, ρ). Rank 1 is best, so a behaviour where more is better shows a negative ρ. A bigger bar means the behaviour tracked the placings more closely on this task, and pilots measured is how much of the analysed field the behaviour applied to — a reading drawn from half the field is thinner than one drawn from all of it. Select a row to see that behaviour plotted against rank — the chart stays in view while you work down the table.

Glide speed between climbs

Each dot is a pilot. ρ = -0.90 (clear pattern, n = 32). More is expected to be better here, and it was: top ranks gather to the right. The curve is a trend fitted through the dots: left to right it runs from about rank 31 to about rank 1. 2 pilots have no value and are not plotted.
  • Field glide speed: median 57.5 km/h · p90 69.3 km/h (32 pilots)
BehaviourStrengthWhat it meansPilots measured
Glide speed between climbs
clear pattern
Gliding wide of the optimal course line
clear pattern
How much of the thermal the pilot climbed before leaving it
clear pattern
How low the pilot gets between climbs
clear pattern
Time spent flying with a gaggle
clear pattern
Distance covered between climbs
clear pattern
Climbs joined on another pilot's marker
clear pattern
Final glide committed to when leaving the last climb
clear pattern
Glide L/D against the field median
clear pattern
Share of race time spent hunting for the next climb
clear pattern
Share of lift turned in that was kept as a climb
some pattern
How often leaving the gaggle paid off
could be chance
Share of the flight spent in air that wasn’t sinking
could be chance
Climb rate at thermal exit
could be chance
Time to core thermals
could be chance
How round and consistent the circles were
could be chance
Share of the height gain made outside thermals
could be chance
Gliding faster when the next climb is stronger
could be chance
Arriving at ESS with height to spare
could be chance
Low saves dug out from the bottom of the band
could be chance
Climbing faster than the pilots sharing the thermal
could be chance

clear pattern is |ρ| ≥ 0.5, some pattern ≥ 0.3 and faint pattern below — each only once the coefficient is bigger than chance alone produces at that many pilots (its noise floor). could be chance (in the statistics: within noise) means shuffling the placings produces a coefficient that size more than 5% of the time, so it cannot be told apart from luck however big it looks. too few pilots is fewer than 8 pilots with a value — not enough to tell either way.

Rank 21 behaviours against one day's results and a few will look strong on luck alone — the ones worth believing are those that repeat across tasks in the competition-level analysis.

Outcome checks

These are not behaviours. They measure the result itself, for example the time behind the leader and the race time lost, so they always follow the places. They are here as a check on the analysis. A weak pattern in this table means that something is wrong in the numbers, and not in the flying of any pilot. Their per-pilot tables stay in the Race craft section below.

OutcomeStrengthWhat it meansPilots measured
Race time behind the leader at ESS
clear pattern
Race time lost against the fastest pilots, leg by leg
clear pattern

The whole field at a glance

1. Steven Crosby
2. Josh Woods
3. Hughbert Alexander
4. Tyler Borradaile
5. Rohan Taylor
6. Paul Bissett-Amess
7. Rory Duncan
8. Peter Burkitt
9. Ward Gunn
10. Rob de Groot
11. Dick Heffer
12. Richard Martin
13. Bradley Elliott
14. Adrian Connor
15. Mick Lamb
16. James Atkinson
17. John Greenwood
19. Peter Garrone
20. Dustan Hansen
21. John Harriott
22. Andy Schmidt
23. James Morgan
24. Col Jackson
25. Brett Gilmour
26. Hossain Tefaili
27. James McGinty
28. Nolan Bear
29. Justin Gilmour
30. Rob Larkin
31. Joe Baker
32. Les Bestt
33. Richard Hughes
34. Andre Johnson
35. Howard Taylor
The pilots in rank order against every behaviour. A darker cell is a better percentile in this field, and an empty cell is a behaviour that does not apply. The columns start with the behaviours whose better end went with better places, continue through the behaviours that separated nobody, and end with the behaviours that ran the other way. A field that one behaviour separated therefore shades dark in the top-left corner, and a field where each pilot won differently does not. The band above rates how much pattern each group of columns holds: a clear, some or faint pattern, noise (could be chance), or too few pilots to tell. The family sections below carry the exact values. † This behaviour has no good or bad direction. The shade is the position in the field, and not the quality.

Pilot style clusters

The groups are flying style, and not score. The spread of ranks in each group shows where that style paid and where it did not. Each group carries the name of its strongest signature. A ★ marks the pilot most typical of their group.

Group A — Committed racers

9 pilots · ranks 1–29 · median 8 · middle half 3–11

  • Low — Share of race time spent hunting for the next climb group median P12 in this field (17 percent) · usually a strength
  • Low — Climb rate at thermal exit group median P13 in this field (-0.0 metres per second)
  • High — How often leaving the gaggle paid off group median P88 in this field (100 percent)
  • Low — Share of the height gain made outside thermals group median P14 in this field (3 percent)
  • 1. Steven Crosby
  • 2. Josh Woods
  • 3. Hughbert Alexander (most typical of this group)
  • 4. Tyler Borradaile
  • 8. Peter Burkitt
  • 10. Rob de Groot
  • 11. Dick Heffer
  • 13. Bradley Elliott
  • 29. Justin Gilmour

Group B — Outclimbed in company

2 pilots · ranks 7–9 · median 8 · middle half 7.5–8.5

  • Low — Climbing faster than the pilots sharing the thermal group median P2 in this field (34 percent) · usually costly
  • High — Share of the height gain made outside thermals group median P98 in this field (32 percent)
  • High — Distance covered between climbs group median P98 in this field (29.4 kilometres) · usually a strength
  • High — Glide L/D against the field median group median P98 in this field (7.92 ratio) · usually a strength
  • 7. Rory Duncan (most typical of this group)
  • 9. Ward Gunn

Group C — High-band operators

6 pilots · ranks 5–23 · median 13 · middle half 7.5–18.5

  • High — How low the pilot gets between climbs group median P87 in this field (57 percent)
  • Low — Time to core thermals group median P17 in this field (41 seconds) · usually a strength
  • High — Glide L/D against the field median group median P79 in this field (1.17 ratio) · usually a strength
  • High — Climb rate at thermal exit group median P77 in this field (0.8 metres per second)
  • 5. Rohan Taylor
  • 6. Paul Bissett-Amess (most typical of this group)
  • 12. Richard Martin
  • 14. Adrian Connor
  • 20. Dustan Hansen
  • 23. James Morgan

Group D — Glide bleeders

10 pilots · ranks 15–28 · median 21.5 · middle half 17.5–24.8

  • Low — Glide L/D against the field median group median P17 in this field (0.84 ratio) · usually costly
  • High — Low saves dug out from the bottom of the band group median P77 in this field (2.0 count)
  • Low — Glide speed between climbs group median P24 in this field (51.7 kilometres per hour) · usually costly
  • Low — Distance covered between climbs group median P25 in this field (1.6 kilometres) · usually costly
  • 15. Mick Lamb
  • 16. James Atkinson
  • 17. John Greenwood
  • 19. Peter Garrone (most typical of this group)
  • 21. John Harriott
  • 22. Andy Schmidt
  • 24. Col Jackson
  • 25. Brett Gilmour
  • 26. Hossain Tefaili
  • 28. Nolan Bear

Group E — Deep diggers

4 pilots · ranks 30–35 · median 32.5 · middle half 30.8–34.3

  • Low — How low the pilot gets between climbs group median P4 in this field (-18 percent)
  • Low — How much of the thermal the pilot climbed before leaving it group median P5 in this field (2 percent)
  • Low — Climbs joined on another pilot's marker group median P5 in this field (0 percent)
  • Low — Time spent flying with a gaggle group median P8 in this field (0 percent)
  • 30. Rob Larkin (most typical of this group)
  • 31. Joe Baker
  • 34. Andre Johnson
  • 35. Howard Taylor

Not clustered: 27. James McGinty — only 5 of 21 metrics available (needs ≥ 60%); 32. Les Bestt — only 7 of 21 metrics available (needs ≥ 60%); 33. Richard Hughes — only 9 of 21 metrics available (needs ≥ 60%).

GlideComp groups the pilots by flying style, and not by score. It transforms the rank of every behavioural metric to a percentile inside the field. It then compares two pilots by the mean percentile gap over the metrics that both pilots have, and never fills in a missing value. Ward-linkage agglomeration forms the groups, and the best mean silhouette selects the number of groups. Each group carries the spread of the GAP ranks of its members, which shows where a style paid and where it did not. On this task, 31 pilots on 21 behavioural metrics formed 5 groups, with k searched from 2 to 6. The mean silhouette is 0.27. A value near 0 means soft group boundaries, and a value near 1 means tight, well-separated groups.

The metrics in detail

best: could be chance (0.32)

best: clear pattern (0.75)

One lane per behaviour, one dot per pilot, placed by percentile within the pilots that behaviour could be measured on — so a dot's position says where a pilot sat, never how spread the field was. Hover a pilot to light them up in every lane at once. Lanes run strongest separator first, the ones that separated nobody last; the table below keeps its usual order. Right is the end the behaviour is expected to be better at — whether it paid on this task is the ranking's question, not this chart's. † No good or bad direction: right is simply the larger value. A lane's count is the pilots it applied to; the rest have no value for it, which is not a score of zero.
#PilotOut-climbCore sLeaveRateKept%TopOut%Round
1Steven Crosby50 (13 shared climbs)100 (13 climbs ≥ 60 s)-0.1 (13 climbs ≥ 90 s)79 (11/14 circling bouts led to climbs)103 (mean on-course altitude 48% of band)0.38 (17 circles, 53% left)
2Josh Woods72 (18 shared climbs)65 (19 climbs ≥ 60 s)-0.0 (14 climbs ≥ 90 s)76 (13/17 circling bouts led to climbs)65 (mean on-course altitude 59% of band)— (7 circles, 57% left)
3Hughbert Alexander65 (13 shared climbs)63 (16 climbs ≥ 60 s)0.2 (15 climbs ≥ 90 s)69 (9/13 circling bouts led to climbs)93 (mean on-course altitude 67% of band)0.28 (18 circles, 56% left)
4Tyler Borradaile49 (18 shared climbs)90 (21 climbs ≥ 60 s)-0.1 (21 climbs ≥ 90 s)86 (18/21 circling bouts led to climbs)74 (mean on-course altitude 45% of band)0.33 (18 circles, 44% left)
5Rohan Taylor78 (19 shared climbs)47 (17 climbs ≥ 60 s)0.9 (14 climbs ≥ 90 s)89 (16/18 circling bouts led to climbs)66 (mean on-course altitude 56% of band)0.20 (128 circles, 91% left)
6Paul Bissett-Amess65 (48 shared climbs)38 (19 climbs ≥ 60 s)1.0 (11 climbs ≥ 90 s)94 (16/17 circling bouts led to climbs)71 (mean on-course altitude 68% of band)0.20 (113 circles, 78% left)
7Rory Duncan37 (5 shared climbs)85 (5 climbs ≥ 60 s)0.5 (5 climbs ≥ 90 s)15 (4/27 circling bouts led to climbs)88 (mean on-course altitude 74% of band)0.30 (21 circles, 71% left)
8Peter Burkitt77 (17 shared climbs)55 (17 climbs ≥ 60 s)0.4 (15 climbs ≥ 90 s)55 (11/20 circling bouts led to climbs)93 (mean on-course altitude 67% of band)0.22 (13 circles, 100% left)
9Ward Gunn32 (5 shared climbs)130 (5 climbs ≥ 60 s)0.4 (5 climbs ≥ 90 s)13 (2/15 circling bouts led to climbs)100 (mean on-course altitude 86% of band)— (8 circles, 63% left)
10Rob de Groot73 (15 shared climbs)64 (23 climbs ≥ 60 s)0.4 (20 climbs ≥ 90 s)67 (16/24 circling bouts led to climbs)60 (mean on-course altitude 40% of band)0.30 (33 circles, 85% left)
11Dick Heffer53 (10 shared climbs)100 (19 climbs ≥ 60 s)-0.3 (19 climbs ≥ 90 s)79 (15/19 circling bouts led to climbs)83 (mean on-course altitude 38% of band)0.36 (25 circles, 32% left)
12Richard Martin77 (35 shared climbs)35 (20 climbs ≥ 60 s)1.5 (13 climbs ≥ 90 s)69 (11/16 circling bouts led to climbs)75 (mean on-course altitude 65% of band)0.19 (197 circles, 52% left)
13Bradley Elliott64 (16 shared climbs)130 (19 climbs ≥ 60 s)-0.2 (19 climbs ≥ 90 s)70 (16/23 circling bouts led to climbs)76 (mean on-course altitude 38% of band)0.35 (15 circles, 73% left)
14Adrian Connor42 (42 shared climbs)41 (33 climbs ≥ 60 s)0.6 (24 climbs ≥ 90 s)80 (20/25 circling bouts led to climbs)72 (mean on-course altitude 58% of band)0.15 (266 circles, 28% left)
15Mick Lamb76 (14 shared climbs)55 (32 climbs ≥ 60 s)0.4 (24 climbs ≥ 90 s)92 (22/24 circling bouts led to climbs)45 (mean on-course altitude 44% of band)0.21 (32 circles, 75% left)
16James Atkinson67 (27 shared climbs)40 (32 climbs ≥ 60 s)0.7 (23 climbs ≥ 90 s)83 (24/29 circling bouts led to climbs)50 (mean on-course altitude 47% of band)0.17 (270 circles, 40% left)
17John Greenwood67 (47 shared climbs)48 (35 climbs ≥ 60 s)0.9 (26 climbs ≥ 90 s)79 (22/28 circling bouts led to climbs)67 (mean on-course altitude 47% of band)0.17 (256 circles, 38% left)
19Peter Garrone74 (21 shared climbs)56 (31 climbs ≥ 60 s)0.8 (23 climbs ≥ 90 s)87 (27/31 circling bouts led to climbs)48 (mean on-course altitude 34% of band)0.18 (300 circles, 94% left)
20Dustan Hansen73 (21 shared climbs)56 (20 climbs ≥ 60 s)0.6 (15 climbs ≥ 90 s)68 (15/22 circling bouts led to climbs)76 (mean on-course altitude 48% of band)0.19 (122 circles, 71% left)
21John Harriott42 (20 shared climbs)60 (30 climbs ≥ 60 s)0.3 (27 climbs ≥ 90 s)56 (15/27 circling bouts led to climbs)54 (mean on-course altitude 44% of band)0.22 (40 circles, 0% left)
22Andy Schmidt73 (34 shared climbs)60 (15 climbs ≥ 60 s)0.9 (11 climbs ≥ 90 s)100 (12/12 circling bouts led to climbs)29 (mean on-course altitude 29% of band)0.17 (135 circles, 64% left)
23James Morgan56 (41 shared climbs)41 (29 climbs ≥ 60 s)0.7 (21 climbs ≥ 90 s)71 (15/21 circling bouts led to climbs)83 (mean on-course altitude 59% of band)0.17 (208 circles, 83% left)
24Col Jackson46 (19 shared climbs)50 (25 climbs ≥ 60 s)0.3 (17 climbs ≥ 90 s)63 (12/19 circling bouts led to climbs)58 (mean on-course altitude 30% of band)0.27 (40 circles, 75% left)
25Brett Gilmour58 (42 shared climbs)43 (25 climbs ≥ 60 s)0.9 (12 climbs ≥ 90 s)73 (8/11 circling bouts led to climbs)31 (mean on-course altitude 28% of band)0.22 (91 circles, 89% left)
26Hossain Tefaili42 (5 shared climbs)150 (11 climbs ≥ 60 s)0.2 (11 climbs ≥ 90 s)77 (10/13 circling bouts led to climbs)31 (mean on-course altitude 26% of band)0.38 (12 circles, 50% left)
27James McGinty———0 (0/23 circling bouts led to climbs)——
28Nolan Bear65 (4 shared climbs)54 (18 climbs ≥ 60 s)0.4 (14 climbs ≥ 90 s)50 (8/16 circling bouts led to climbs)58 (mean on-course altitude 50% of band)0.20 (51 circles, 49% left)
29Justin Gilmour73 (15 shared climbs)58 (14 climbs ≥ 60 s)0.3 (13 climbs ≥ 90 s)73 (11/15 circling bouts led to climbs)64 (mean on-course altitude 52% of band)0.21 (22 circles, 27% left)
30Rob Larkin66 (6 shared climbs)40 (4 climbs ≥ 60 s)0.7 (3 climbs ≥ 90 s)56 (5/9 circling bouts led to climbs)-10 (mean on-course altitude -7% of band)0.23 (54 circles, 98% left)
31Joe Baker56 (27 shared climbs)6 (4 climbs ≥ 60 s)2.0 (2 climbs ≥ 90 s)40 (2/5 circling bouts led to climbs)25 (mean on-course altitude 18% of band)0.22 (53 circles, 77% left)
32Les Bestt65 (2 shared climbs)85 (2 climbs ≥ 60 s)0.5 (2 climbs ≥ 90 s)——— (3 circles, 0% left)
33Richard Hughes—80 (6 climbs ≥ 60 s)0.4 (6 climbs ≥ 90 s)—41 (mean on-course altitude 10% of band)— (3 circles, 100% left)
34Andre Johnson58 (4 shared climbs)53 (10 climbs ≥ 60 s)0.5 (7 climbs ≥ 90 s)60 (6/10 circling bouts led to climbs)-7 (mean on-course altitude -13% of band)0.33 (12 circles, 58% left)
35Howard Taylor74 (11 shared climbs)53 (5 climbs ≥ 60 s)1.4 (4 climbs ≥ 90 s)33 (1/3 circling bouts led to climbs)11 (mean on-course altitude 4% of band)0.15 (73 circles, 86% left)

Share of lift turned in that was kept as a climb

Measured in percent · no expected direction

How selective the pilot is about the lift they stop for. Each period of circling of 30 s or more after the start counts as lift that the pilot sampled. If the period overlaps a detected thermal, the pilot kept that lift. If it does not, they turned a few circles and left it. The value is the percentage kept. A low value means they are selective. A high value means they keep almost every climb they turn in. There is no expected direction: selection wins on a strong day and wastes time on a weak one.

Acceptance by hour

HourMedian accepted (%)pilots
10013
7630
7327
6019
3710
01

Median per-pilot acceptance %, bucketed by the hour (competition time zone).

How round and consistent the circles were

Measured in ratio · lower is better

Whether the pilot flies clean, repeatable circles, or moves around the thermal. We fit each detected circle by least squares. The RMS fit error divided by the fitted radius measures how round the turn was. The value is the median over all of the circles of the pilot. A lower value means smoother and more consistent turns.

Turn direction across the field: 63% left (2656 circles).

best: clear pattern (0.90)

One lane per behaviour, one dot per pilot, placed by percentile within the pilots that behaviour could be measured on — so a dot's position says where a pilot sat, never how spread the field was. Hover a pilot to light them up in every lane at once. Lanes run strongest separator first, the ones that separated nobody last; the table below keeps its usual order. Right is the end the behaviour is expected to be better at — whether it paid on this task is the ranking's question, not this chart's. † No good or bad direction: right is simply the larger value. A lane's count is the pilots it applied to; the rest have no value for it, which is not a score of zero.
#PilotGlideSpdGlideL/DSpeedToFlyWide%Dolphin%
1Steven Crosby72.1 (8 glides, 52 min gliding)1.16 (3 legs compared)-0.9 (7 glide→climb pairs)11 (3 legs completed)6 (311 of 5555 m gained outside thermals)
2Josh Woods72.6 (12 glides, 55 min gliding)0.94 (3 legs compared)-10.9 (11 glide→climb pairs)13 (3 legs completed)3 (190 of 6212 m gained outside thermals)
3Hughbert Alexander69.4 (6 glides, 56 min gliding)1.09 (3 legs compared)-4.4 (5 glide→climb pairs)7 (3 legs completed)2 (102 of 5021 m gained outside thermals)
4Tyler Borradaile69.8 (8 glides, 50 min gliding)1.13 (3 legs compared)12.4 (7 glide→climb pairs)13 (3 legs completed)5 (268 of 5352 m gained outside thermals)
5Rohan Taylor67.8 (16 glides, 68 min gliding)1.34 (3 legs compared)1.5 (15 glide→climb pairs)14 (3 legs completed)5 (228 of 4345 m gained outside thermals)
6Paul Bissett-Amess64.1 (24 glides, 72 min gliding)1.20 (3 legs compared)1.1 (23 glide→climb pairs)23 (3 legs completed)13 (627 of 4731 m gained outside thermals)
7Rory Duncan68.9 (2 glides, 34 min gliding)9.41 (3 legs compared)—7 (3 legs completed)26 (258 of 994 m gained outside thermals)
8Peter Burkitt59.7 (10 glides, 68 min gliding)0.98 (3 legs compared)1.2 (9 glide→climb pairs)9 (3 legs completed)3 (176 of 5252 m gained outside thermals)
9Ward Gunn61.9 (2 glides, 106 min gliding)6.42 (3 legs compared)—14 (3 legs completed)39 (506 of 1302 m gained outside thermals)
10Rob de Groot64.6 (18 glides, 66 min gliding)0.86 (3 legs compared)-0.8 (17 glide→climb pairs)8 (3 legs completed)2 (159 of 6626 m gained outside thermals)
11Dick Heffer65.8 (9 glides, 67 min gliding)0.94 (3 legs compared)-3.2 (8 glide→climb pairs)20 (3 legs completed)2 (181 of 7587 m gained outside thermals)
12Richard Martin58.5 (27 glides, 100 min gliding)1.06 (3 legs compared)0.1 (26 glide→climb pairs)31 (3 legs completed)13 (882 of 6914 m gained outside thermals)
13Bradley Elliott63.7 (13 glides, 72 min gliding)0.96 (3 legs compared)-8.9 (12 glide→climb pairs)19 (3 legs completed)4 (298 of 7755 m gained outside thermals)
14Adrian Connor57.2 (25 glides, 93 min gliding)1.00 (3 legs compared)0.4 (24 glide→climb pairs)18 (3 legs completed)7 (435 of 5835 m gained outside thermals)
15Mick Lamb56.7 (19 glides, 86 min gliding)0.94 (3 legs compared)0.9 (18 glide→climb pairs)25 (3 legs completed)7 (489 of 7501 m gained outside thermals)
16James Atkinson56.5 (29 glides, 91 min gliding)0.97 (3 legs compared)-0.3 (28 glide→climb pairs)30 (3 legs completed)6 (409 of 6752 m gained outside thermals)
17John Greenwood64.0 (29 glides, 84 min gliding)0.78 (3 legs compared)8.9 (28 glide→climb pairs)31 (3 legs completed)16 (1416 of 9103 m gained outside thermals)
19Peter Garrone50.7 (28 glides, 85 min gliding)0.68 (2 legs compared)1.8 (27 glide→climb pairs)20 (2 legs completed)4 (318 of 7400 m gained outside thermals)
20Dustan Hansen60.3 (20 glides, 82 min gliding)1.24 (2 legs compared)3.0 (19 glide→climb pairs)32 (2 legs completed)4 (257 of 6305 m gained outside thermals)
21John Harriott51.6 (16 glides, 95 min gliding)0.84 (1 leg compared)2.8 (15 glide→climb pairs)41 (1 leg completed)7 (393 of 6040 m gained outside thermals)
22Andy Schmidt57.8 (14 glides, 59 min gliding)0.89 (1 leg compared)4.7 (13 glide→climb pairs)45 (1 leg completed)20 (942 of 4751 m gained outside thermals)
23James Morgan56.1 (22 glides, 79 min gliding)1.14 (1 leg compared)-0.0 (21 glide→climb pairs)38 (1 leg completed)9 (442 of 4949 m gained outside thermals)
24Col Jackson43.7 (10 glides, 53 min gliding)0.59 (1 leg compared)4.6 (9 glide→climb pairs)27 (1 leg completed)4 (174 of 4510 m gained outside thermals)
25Brett Gilmour51.1 (16 glides, 54 min gliding)0.92 (1 leg compared)0.3 (15 glide→climb pairs)68 (1 leg completed)15 (513 of 3315 m gained outside thermals)
26Hossain Tefaili51.8 (5 glides, 27 min gliding)0.77 (1 leg compared)2.0 (4 glide→climb pairs)25 (1 leg completed)4 (118 of 2912 m gained outside thermals)
27James McGinty—————
28Nolan Bear44.4 (8 glides, 53 min gliding)—1.0 (7 glide→climb pairs)—5 (180 of 3816 m gained outside thermals)
29Justin Gilmour53.7 (8 glides, 34 min gliding)—-4.9 (7 glide→climb pairs)—2 (69 of 2989 m gained outside thermals)
30Rob Larkin56.0 (6 glides, 22 min gliding)—-6.2 (5 glide→climb pairs)—25 (297 of 1209 m gained outside thermals)
31Joe Baker49.1 (3 glides, 17 min gliding)————
32Les Bestt—————
33Richard Hughes42.5 (1 glides, 10 min gliding)————
34Andre Johnson43.4 (6 glides, 21 min gliding)—-2.2 (5 glide→climb pairs)—13 (133 of 1021 m gained outside thermals)
35Howard Taylor52.6 (3 glides, 10 min gliding)————

Glide speed between climbs

Measured in kilometres per hour · higher is better

How fast the pilot moves down the course when they are on a glide. The value is the duration-weighted mean ground speed over every glide after the start, which is the glide distance divided by the glide time. A higher value means more ground covered in each minute between climbs.

Field glide speed: median 57.5 km/h · p90 69.3 km/h (32 pilots)

best: clear pattern (0.67)

best: clear pattern (0.66)

best: clear pattern (0.59)

Footnotes

1 pilot in the standings but not in this analysis

  • Nariyuki Watabe — scored from a manual flight report — no tracklog to analyse

The correlations are measured against the published ranks, and those ranks include these pilots. Their behaviour cannot be measured without a tracklog.

How the field is compared

Everything that compares pilots to each other uses one shared clock. That includes gaggles, shared thermals, and the position of each pilot at the same moment. GlideComp resamples every track onto a common 10-second grid. Two pilots are therefore always compared at the same instant, whatever rate their instruments logged at.

Metric glossary

How GlideComp measures every metric on this page. On screen, the ⓘ beside a metric opens the same description in place. On paper, this section is the reference for all of them.

Day profile & wind

The day’s wind, hour by hour and leg by leg(“Wind” in tables)
Measured in kilometres per hour · no expected direction

What the air did, read from the field itself. We estimate the wind from the circling of every pilot. The first method is the drift of the circle centre, and the second method, used when the first is not available, is the modulation of the ground speed. We then average the vectors two ways. The table by hour of day shows how the wind increased and changed direction through the day. The table by speed-section leg shows the wind on each part of the course. This metric describes the day, so it has no value for each pilot.

How strong the day’s climbs were, hour by hour(“Climb/hr” in tables)
Measured in metres per second · no expected direction

When the day started, reached its peak, and ended. We group the thermal climbs of all pilots by the hour in which each climb started, labelled in the time zone of the competition. The median and the 90th-percentile average climb rate for each hour show how the lift developed. This metric describes the day, so it has no value for each pilot.

Share of the flight spent in air that wasn’t sinking(“NonSink%” in tables)
Measured in percent · no expected direction

How much of the flight was in air worth being in. The value is the share of the airborne time of a pilot, on the shared grid, with a 30 s-smoothed vario at or above −0.5 m/s. The time they flew, the line they steered and the way the flight ended all feed this value. It is therefore a reading of the day as much as of the pilot. There is no expected direction, and the sign of the correlation is the finding. The timing table compares the window of the day’s best climbs against the time when the field launched.

Climbing

Climbing faster than the pilots sharing the thermal(“Out-climb” in tables)
Measured in percent · higher is better

When this pilot and other pilots were in the SAME thermal, who climbed faster? In every thermal that two pilots or more used, we rank each use by its average climb rate. The percentile of a use is the share of uses that were strictly slower. The value is the duration-weighted mean percentile over the shared climbs of the pilot. 50% is exactly average. 80% means they climbed faster than four in five of the pilots they shared lift with. The shared thermal is what separates centring skill from thermal selection: a pilot who only found better air gets no higher value here.

Time to core thermals(“Core s” in tables)
Measured in seconds · lower is better

How long the pilot takes to get into the best lift after they arrive in a thermal. For each thermal of 60 s or more, we measure the seconds from the entry until the 30 s rolling climb rate first reaches 90% of its peak in that thermal. The value is the median across the thermals of the pilot. Every second here is a second spent climbing slower than the thermal can carry them.

Climb rate at thermal exit(“LeaveRate” in tables)
Measured in metres per second · no expected direction

The median climb rate that the pilot left thermals at. For each thermal of 90 s or more, we take the climb rate over its final 30 s. A high value means they leave lift that still works. A low value means they stay in a climb until nothing is left. This is an absolute rate, so read it against the day: compare it with the median climb in "How strong the day’s climbs were". A pilot who leaves at 1.5 m/s leaves a good climb on a 1 m/s day, and takes the worst lift available on a 4 m/s day. There is no expected direction. The sign of the correlation says which behaviour paid on this task.

Share of lift turned in that was kept as a climb(“Kept%” in tables)
Measured in percent · no expected direction

How selective the pilot is about the lift they stop for. Each period of circling of 30 s or more after the start counts as lift that the pilot sampled. If the period overlaps a detected thermal, the pilot kept that lift. If it does not, they turned a few circles and left it. The value is the percentage kept. A low value means they are selective. A high value means they keep almost every climb they turn in. There is no expected direction: selection wins on a strong day and wastes time on a weak one.

How much of the thermal the pilot climbed before leaving it(“TopOut%” in tables)
Measured in percent · no expected direction

Does the pilot climb to the top of every thermal, or leave with lift still above them? We take the altitude where they left each thermal after the start, as a percentage of the day’s working band. 0% is the floor of the field and 100% is its ceiling. The value is the median. There is no expected direction: a climb to the top buys height in reserve, and an early departure buys time.

How round and consistent the circles were(“Round” in tables)
Measured in ratio · lower is better

Whether the pilot flies clean, repeatable circles, or moves around the thermal. We fit each detected circle by least squares. The RMS fit error divided by the fitted radius measures how round the turn was. The value is the median over all of the circles of the pilot. A lower value means smoother and more consistent turns.

Gliding

Glide speed between climbs(“GlideSpd” in tables)
Measured in kilometres per hour · higher is better

How fast the pilot moves down the course when they are on a glide. The value is the duration-weighted mean ground speed over every glide after the start, which is the glide distance divided by the glide time. A higher value means more ground covered in each minute between climbs.

Glide L/D against the field median(“GlideL/D” in tables)
Measured in ratio · higher is better

Whether the pilot found better air on glide than the other pilots on the same leg. For each completed speed-section leg, we take the pilot's glide-phase L/D. That is the path distance divided by the net altitude lost during the glides, and we skip a leg that loses less than 100 m. We divide it by the median L/D of the field on that same leg, and then average over the legs. 1.10 means the pilot glided 10% further for each metre lost than the usual pilot on those legs.

Gliding faster when the next climb is stronger(“SpeedToFly” in tables)
Measured in kilometres per hour · higher is better

Speed to fly: the pilot flies faster when a good climb is in front of them, and slower when it is not. We pair each glide after the start with the climb rate of the next thermal that starts within 5 minutes. The value is the mean glide speed before climbs stronger than the median, minus the mean glide speed before weaker climbs. +8 km/h means the pilot flew 8 km/h faster into the good climbs. This is a PROXY, and not true speed to fly, because there is no glider polar data.

Gliding wide of the optimal course line(“Wide%” in tables)
Measured in percent · lower is better

How much further the pilot flew on glide than the optimised course line needed. 0% is a flight exactly along the line, and 12% is a glide 12% further than necessary. On each completed speed-section leg, we compare the pilot's route with the optimised distance of the leg, weighted by that optimised distance. Only the glides are measured at their full path length. Circling and searching contribute their entry-to-exit displacement instead. A climb or a search for lift therefore never reads as a wide line, because a pilot chooses a line only on glide. 0% is a real value that a pilot can reach: a pilot who flies the line of the optimiser scores exactly zero.

Share of the height gain made outside thermals(“Dolphin%” in tables)
Measured in percent · no expected direction

Dolphin flying: how much of the height that the pilot gained came outside of circling. The value is the share of the altitude gain after the start, smoothed over 10 s, that the pilot made outside a detected thermal. There is no expected direction. The sign of the correlation shows whether dolphin flying paid on this day.

Decision-making

How low the pilot gets between climbs(“Floor%” in tables)
Measured in percent · no expected direction

How low the pilot goes before the next climb. A high value is a race with height in reserve, and a low value is a flight that goes down near the ground. We take each pair of climbs that the pilot made after the start, and we find the lowest point between them. We keep only the gaps that go down 100 m or more, because a top-up between two climbs is not a descent. We do not count a sled run or the glide to goal, because the pilot made no climb after them. The value is the median of those low points, as a percentage of the day's working band. 0% is where the lowest tenth of the field's climbs started, and 100% is where the highest tenth stopped. Thus a negative value shows that the pilot went lower than almost all of the field. The pilot must have two or more of these descents. There is no expected direction. The sign of the correlation says whether height in reserve pays.

Low saves dug out from the bottom of the band(“LowSaves” in tables)
Measured in count · no expected direction

How many times the pilot got low and climbed out again. We count the climbs after the start that the pilot entered below 15% of the working band, and that then gained 300 m or more. Those are true low saves. Zero is a real value, and not a missing one: it means the pilot never got that low. There is no expected direction. The sign of the correlation says whether a climb-out or a flight that stays high pays.

Distance covered between climbs(“km/climb” in tables)
Measured in kilometres · higher is better

How far the pilot gets down the course before they must stop and circle again. This is the direct reading of how often they stop. The value is the scored flown distance divided by the number of thermals taken after the start, so 3 km means three kilometres of course for each climb. The pilot must fly 20 km or more. The note of each pilot adds their mean climb percentile inside shared thermals, so you can read the number of stops together with the climb strength. Long legs between weak climbs is a different day from long legs between strong ones.

Share of race time spent hunting for the next climb(“Search%” in tables)
Measured in percent · lower is better

Time that goes into neither a climb nor progress down the course. This is the time spent to find lift, to stay up, and to decide what to do next. The value is the share of the speed-section time, from the start to ESS or to the landing, in which the pilot neither climbed in a thermal nor glided with real net speed. A lower value means less time lost between climbs.

Gaggle

Time spent flying with a gaggle(“InGaggle%” in tables)
Measured in percent · no expected direction

Whether the pilot raced with other pilots or alone. The value is the share of their flying time after the start inside a detected gaggle, that is, clustered with one other racing pilot or more on the shared time grid. There is no expected direction. A gaggle increases the power to search for lift, but it also holds a pilot to its own speed. The sign of the correlation says which of the two occurred here.

Climbs joined on another pilot's marker(“Marked%” in tables)
Measured in percent · no expected direction

How much of the lift of the pilot another pilot found first. The value is the share of their climbs after the start where another pilot was already established in the same thermal when they arrived. Established means 30 s or more into the climb, and still climbing. A high value means they mostly climb on the markers of other pilots. A low value means they find their own air. There is no expected direction. A marker is free information, but it puts a pilot where the last climb was, and not where the next one is.

How often leaving the gaggle paid off(“LeaveWin%” in tables)
Measured in percent · no expected direction

When a pilot leaves a gaggle that continues to fly, did the departure pay off? We compare the arrival of the pilot who left at the next turnpoint against the median arrival of the pilots who stayed. A win rate of more than 50% means their departures beat the gaggle. A pilot counts as a pilot who stayed only if they were still in the gaggle after the split, and reached that turnpoint after it.

Race craft

How long after the gate opened the pilot started(“StartDly” in tables)
Measured in seconds · lower is better

Every second between the opening of the gate and the crossing of the start line is a second lost for nothing. The value is the seconds from the start gate taken to the scored SSS crossing. On an elapsed-time task, the pilot’s own crossing is the reference, so the delay is 0 by definition. The start table adds the crossing altitude, and the distance behind the leading pilot who had already started.

Race time lost against the fastest pilots, leg by leg(“TimeLost” in tables)
Measured in seconds · lower is better

For each completed speed-section leg, we compare the leg time of the pilot with the mean of the top 10 pilots by rank who completed that leg. Only the losses count, and we add them together. The sum of the leg times is the race time, and the rank defines the reference, so this metric follows the result by construction. Read the waterfall table, which shows every leg against the task winner, for the diagnosis. Do not read the correlation as a finding.

Race time behind the leader at ESS(“Behind” in tables)
Measured in minutes · lower is better

At each speed-section turnpoint, we compare the elapsed race time of the pilot, which is the reaching time minus their own start, with the fastest pilot to that turnpoint. The value is the minutes behind at ESS. It follows the final rank almost exactly, because this metric is the sanity check of the evaluation.

Arriving at ESS with height to spare(“Spare m” in tables)
Measured in metres · lower is better

Height still available at ESS that the pilot no longer needed. That altitude was available for more speed, and the pilot did not use it. The value is the altitude at ESS minus the altitude needed to glide to goal at the standard glide ratio of the sport, which is 5.0 for HG and 4.0 for PG (S7F §12.3.6). A large positive margin means the pilot arrived too high. A margin near zero means they flew the final glide with little height to spare.

Final glide committed to when leaving the last climb(“FinalGl” in tables)
Measured in ratio · no expected direction

How optimistic the pilot was about their final glide. A pilot wins or loses a task by the height at which they leave the last climb. At the last climb of the pilot before ESS, or before the landing, we divide the distance to goal by their height above goal. That is the glide ratio they committed to. 8 means they left and needed 8:1 to make goal. The value counts only when that climb ended within 1.5 times the distance of the final leg from goal. There is no expected direction: a marginal glide wins if it connects, and loses if it does not.