Real engine on real G3 (RCYC 6/15, f2a2cacc) · headline KB-strict recovery (s) · 2 of 3 boats = heading-suspect → SOG metrics trusted, TWA/count muted · benchmark bands re-anchored to KB-strict · REAL DATA, engine-validated
‹RCYC · R1JUN 15 ’253/3
MANEUVERSKB recovery · quality-gated
OverviewTacksGybesLog± Est. loss off
181Innocent B.GPS COURSE298ChimichuriGPS COURSE441AlchemistHEADING OK
⚑ Data quality: 181 & 298 carry a heading sensor ~115–121° off GPS course (40% reversed) — the engine uses GPS course for these boats. Recovery seconds (from speed) are trusted; maneuver count, tack angle & TWA are approximate ·gps.
MANEUVER OVERVIEWrace
Maneuvers T+G
6 2+4
10 3+7
8 4+4
Avg recovery s · SOG
13.5
15.8
14.7
Major moments ⚑
0
1gybe L4 · never regained 6.7 kn
2gybe L2 · tack L3
Back to entry <20 s %
83
70
75
recovery = seconds from turn start until SOG regains entry speed (marked ~3 s before turn) · >30 s ⇒ times out as a MAJOR MOMENT ⚑ · bands ≤12 s strong · 12–20 typical · >30 blown
Even the fleet leader blew two maneuvers. 441 (clean heading) has the best avg recovery but timed out twice — a gybe on Leg 2 and a tack on Leg 3 that never got back to entry speed. The old engine metric hid both as "5.5 s / 12 s recovered." That's the coaching signal.
TACKSbeats · 441 clean
Count
2
3
4
Avg recovery s
14.0
16.5
15.0
Best
13 sL1
13 sL1
12 s11:26 · L1
Worst
15 sL3
20 sL3
⚑ timeoutL3 · never regained 6.1 kn
441's Leg-3 tack is the debrief clip — entered at 6.1 kn, dropped to 4.0, and never rebuilt to entry within 30 s. Tap it for the recovery curve and a jump to replay.
GYBESruns · 441 clean
Count
4
7
4
Avg recovery s
13.0
15.2
14.3
Best
11 sL4
12 sL2
12 s11:34 · L2
Worst
16 sL2
⚑ timeoutL4
⚑ timeoutL2 · 6.4 kn
298's Leg-4 gybe is its race-defining moment — entered at 6.7 kn on the final run and never recovered (⚑). The engine reported "8.3 s recovered." KB-strict calls it what it is.
441 MANEUVERSKB-strict · real · by leg
LEG 1 ↑ BEAT · 2 TACKS
11:26:40441⤸ P→S14 s100%
RECOVERY CURVE · ENTRY 5.4 kn (SOG, ~3 s BEFORE TURN START) · SOG-BASED, TRUSTED ON THIS BOAT
duration 7 s · low 3.5 kn · 80% @ 6 s · 95% @ 11 s · 100% @ 14 s
vmg in→out 4.9 → 4.1 knTWA exit 41°tack angle 78°dist in turn 46 mSOG −20 s 5.6 knSOG +20 s 5.2 kn▸ jump to replay · 11:26:33–11:26:47
11:31:12441⤹ S→P12 s100%
LEG 2 ↓ RUN · 2 GYBES
11:36:08441⤸ gybe⚑ >30 sMOMENT
MAJOR MOMENT — NEVER REGAINED ENTRY SPEED (6.4 kn) ON THE NEW GYBE
duration 6 s · low 4.9 kn · peaked at 91% of entry, then faded · engine-current called this "5.5 s recovered" · jump to replay ›
11:38:44441⤹ gybe12 s100%
LEG 3 ↑ BEAT · 2 TACKS
11:44:20441⤹ S→P19 s100%
11:47:55441⤸ P→S⚑ >30 sMOMENT
LEG 4 ↓ — 2 more gybes (13 s · 18 s) · tap to expand · switch boat ▸ 298 / 181 (GPS course)
Why the log shows 441 first: it's the only clean-heading boat, so its per-maneuver detail is precise. 298 & 181 render with the GPS course treatment — recovery seconds still trustworthy (speed-based), but tack angle / TWA-exit are approximate. "± Est. loss" (off) needs steady-wind data — a well-timed tack on a shift is a gain.
☰ MENU
Map◫Data
v0.44 · real
iPhone · 390pt · full scroll · REAL G3 data
WHAT THE EVIDENCE CHANGED
Per-boat quality is now a primary layer. On the real 3-boat sample, 2 of 3 boats have suspect heading (181: 115°/40% reversed · 298: 121°/40%). N2 finding. Each carries a GPS COURSE badge; 441 gets HEADING OK.
Trust hierarchy → visual hierarchy. Recovery seconds use SOG only, so they're trusted on every boat and render normally. Count / tack-angle / TWA-exit come from heading → muted ·gps on suspect boats. The headline metric survives bad data.
Recovery numbers are bigger than v4a assumed. Real KB-strict on 441: 12–19 s, two timeouts of 8. Bands re-anchored: ≤12 strong · 12–20 typical · >30 blown. (v4a's "5–7 s good" was the wrong scale — that was the 0.9× engine metric.)
Major moments are the payload. Engine-current hid the two worst maneuvers of the race. 441 timed out twice; 298's Leg-4 gybe never recovered. The ⚑ count + flagged rows are the reason to open the page.
Detail keeps ALL fields (your memo): vmg in/out, TWA exit, tack angle, dist-in-turn, decay 80/95/100, SOG ±20 s, and the turnPhase ISO window as a replay clip — one tap down.
HONESTY, NOT FABRICATION
Every number here is from the real engine on real tracks (G3 f2a2cacc). 441's maneuvers are exact; 298/181 counts are cog-derived (post-N2), shown with the badge — the design tells the truth about precision.
Depends on N1 (radius fix) for leg grouping and N2 (heading gate) for which boats to trust — both are the first two backlog items. The page is built to render honestly before AND after those land.