A helmsman heats up to build apparent wind, then dives to a deeper angle while the boat is still carrying that speed. It's a real, repeating cycle in S/V Chim Chim's raw telemetry — and it may be why the fleet's pooled polar data makes sustained deep-angle sailing look faster than it actually is.
polar_coords_race, groupID 1 (TransPac ’15), isTransition=0, TWA 90–175°. Full method in the footer.The theory, from the helm: heat up to a hotter angle to build apparent wind and boatspeed, then bear away hard while that speed is still carried forward — the boat goes deep and stays fast, for a while, before it settles back down. Below is 3½ minutes of continuous, unedited raw telemetry (no averaging window, no cherry-picked segment boundaries) from a steady trade-wind stretch of TransPac ’15. It's doing exactly that, repeatedly.
Callout cycle (t = 104–130s): TWA bottoms out at 139.6° — the heat-up. Boatspeed keeps climbing for another 6 seconds, peaking at 11.97kt. Only then does TWA swing out to its deepest point, 152.0°, still carrying most of that speed — before bleeding off 1.9kt (−16%) over the next 20 seconds spent holding the angle. Same shape repeats at least three more times in this one 228-second window.
To check this wasn't just eyeballing one pretty chart, I cross-correlated TWA against boatspeed across a clean, gap-free 6-hour single-tack stretch of the same race (detrended against a 10-minute rolling mean, so only the fast oscillation survives). If the heat/dive story is real, boatspeed should lag a hot angle by some seconds, and the dive should lag peak boatspeed by a further few seconds.
Two lags dominate, and they run in the sequence the theory predicts. At lag +15s, a hot TWA now anti-correlates with boatspeed 15 seconds later (r = −0.50) — heat up, and speed follows a few seconds behind. At lag −10s, boatspeed 10 seconds ago correlates with a deep TWA right now (r = +0.28) — the dive comes after the speed, not before it. The concurrent correlation (lag 0) is near zero — by the moment the boat is actually pointed deep, the causal work already happened a few seconds earlier.
The direct test: find every excursion into TWA ≥ 145° across the whole race — 2,303 of them — and group them by how long the boat actually held that angle before coming back up. If the deep-angle speed is a real, sustainable capability, duration shouldn't matter much. If it's borrowed momentum from the heat-up, it should decay the longer the angle is held.
| Duration held | Excursions | Samples | Mean Bsp | Mean Tws | Bsp / Tws |
|---|---|---|---|---|---|
| ≤15s | 1,821 | 2,995 | 12.92kt | 15.5kt | 0.83 |
| 15–30s | 363 | 1,704 | 11.86kt | 14.6kt | 0.82 |
| 30–60s | 105 | 904 | 10.45kt | 13.4kt | 0.78 |
| 60–120s | 12 | 189 | 8.06kt | 11.6kt | 0.70 |
| 120–300s | 2 | 59 | 8.42kt | 14.8kt | 0.57 |
Speed drops with every step out — and it's not just wind easing: the wind-normalized ratio (Bsp / Tws) falls too, from 0.83 down to 0.70 across the well-sampled buckets. 80% of every TWA ≥ 145° sample in the race (4,699 of 5,851) comes from an excursion lasting 30 seconds or less — the acceleration phase, not a held angle. The last bucket (n = 2) is too thin to trust on its own, but doesn't break the trend either.
BTF's pooled empirical VMG curve — the one behind the 152.5° optimal angle and the “sail 145–155°, gain 24% wind-axis VMG” finding — is built by binning 5-second polar_coords_race samples straight by raw TWA, with no filter for how long that angle was actually held. This maneuver is exactly what fills the 145–165° bins: brief, speed-carrying dives, not sustained sailing. The genuinely-held samples in this same dataset run 10–35% slower than the bin average suggests.
The empirical curve is therefore likely real in direction, overstated in magnitude — deep angles may still beat shallow ones, but probably by less than 24%, and the number isn't necessarily achievable by steering to a fixed deep heading and holding it. The “Claude.ai Analysis” landing page on btf.chimchim.com was pulled from site navigation pending a dwell-time-filtered rebuild of the curve. This report doesn't yet supply that corrected number — it establishes that one is needed. Section 05 below complicates this further: the bias doesn't look uniform across wind strength, so the eventual fix likely needs to run per wind band, not once on the pooled set.
TransPac ’15 sailed in 11–15kt trade wind. TransPac 2017 (groupID 33) sailed the same ocean crossing two years later in genuinely heavier, gustier breeze — 17–25kt, averaging 19kt in the clean window tested below. Same boat, same three checks, rerun end to end.
Same shape, different center of gravity. TP’15's curve is a saddle: near zero at lag 0, with its real peaks pushed out to ±10–15s — the signature of speed being carried from one moment into another. TP’17's curve is a single sharp spike sitting right on lag 0 (r = +0.44, and it holds at 0.47 after partialling out wind-strength fluctuation) — deep and fast happening together, not offset. That's not the momentum-carry signature; it looks like a direct, real-time relationship between angle and speed in strong wind.
TP’15 (copper) loses 4.9kt (−38%) going from a brief dip to a minute-plus hold. TP’17 (teal) loses only 1.0kt (−7%) over the same range — nearly flat. Whatever's happening at 19kt of breeze, it isn't the same borrowed-momentum story.
The helm behavior replicates almost exactly — TP’17 sailors still heat up before diving 95.4% of the time (8.0° on average, vs TP’15's 10.2°), and if anything the compositional skew is worse (86.9% of deep samples come from excursions ≤30s, vs TP’15's 80%). What doesn't replicate is the speed consequence. In TP’15's 11–15kt air, that heat-and-dive habit is doing real work — it's the only way to get a boatspeed number that high at that angle, and it fades fast once the angle is held. In TP’17's 17–25kt air, boats that hold a deep angle for a minute are running almost as fast as boats that just dipped into it — deep genuinely seems to go with fast there, independent of the gust (the relationship survives controlling for Tws fluctuation, partial r = 0.47). That's consistent with an earlier BTF finding that the crossover angle where real performance overtakes the VPP's target deepens as wind builds (117° at 10kt → 148° at 25kt, pooled fleet and both TransPacs independently) — a real aerodynamic story, not a telemetry artifact.
Net so far: the Hookup Maneuver is real everywhere as a helm habit, but its bias on the pooled polar looks concentrated in lighter-to-moderate wind. A third, independent race — below — sharpens that.
SD2PV 2018 (groupID 120, 1,205nm, 5 days) is a shorter, coastal-to-offshore race — a different course shape than either TransPac, and its wind ranged much wider over the five days (5–32kt, mean 15kt) than either trade-wind crossing. That range turned out to matter: the naive whole-race duration-bucket test came back looking almost flat, because longer deep-angle holds in this race happen to coincide with the wind building over the following days — a real confound the TransPac tests didn't have to worry about, since their wind stayed comparatively narrow. Restricting to the 10–18kt band (806 of 1,409 deep excursions, the range that overlaps TP’15) removes it.
SD2PV's curve (violet) lands almost on top of TP’15's (white) — the same near-zero concurrent correlation with real peaks pushed to ±10s. Two races, different courses, different years, same 11–15kt-ish wind band, same signature. TP’17's single spike at lag 0 stands alone as the outlier — and that outlier is the strong-wind race.
Three races, three decay rates, ranked exactly by wind strength: TP’15 (11–15kt) falls 38%, SD2PV (10–18kt band) falls 10%, TP’17 (17–25kt) falls 7%. The 120–300s bucket is dropped here — too thin in all three races (n = 1–4) to plot honestly.
Two independent races in a similar wind band produced the same signature; a third, much windier race didn't. That's a real pattern, not a TransPac-specific quirk of one fleet or one course. It also sharpens the fix: the dwell-time-filtered VMG rebuild this report keeps pointing to should be run per wind band, and the correction will matter most in the moderate air where most of the fleet's downwind sailing actually happens — not in the rare strong-breeze legs where deep angles may already be honestly fast.
Coastal Cup 2018 (groupID 143, Monterey→Santa Barbara, 295nm, ~1.5 days) is the shortest race tested yet, and a genuinely different animal — a coastal course, not an ocean crossing, sailed mostly in 11–19kt. Its wind (mean 15.7kt in the cleanest window) sits almost exactly between TP’15/SD2PV's ~13kt and TP’17's ~19kt. So does its result.
Coastal Cup almost never holds a deep angle at all — 97.1% of its TWA ≥ 145° samples come from excursions of 30 seconds or less, the thinnest of any race tested (only 3 excursions in the whole race made it past 30 seconds, none past 60). Read that as data in itself: a buoy-to-buoy coastal course with more maneuvering and fewer multi-hour steady runs doesn't give the boat the chance to sustain a deep angle the way an open-ocean trade-wind crossing does — so it's excluded from the duration-decay comparison chart above, but its cross-correlation lag structure is still fully measurable, and it's the most interesting result yet.
Six races, six different courses, five different years — and the concurrent TWA↔boatspeed correlation rises with mean wind speed. TP’15, SD2PV, N2E, and SB2KH cluster near or just below zero across 11–14kt (N2E actually the most negative of the six, at −0.10); Coastal Cup and TP’17 sit clearly higher at 15.7 and 18.8kt. This looks like a continuous transition, not a hard cutoff: somewhere through the mid–high teens of wind, the balance shifts from “deep is fast because you carried speed into it” to “deep is fast right now, on its own.” Six points isn't enough to fit a precise crossover speed, but the direction and the low-wind cluster are consistent across every independent test in this report.
This is the clearest version of the finding so far. The Hookup Maneuver's distortion of the pooled polar isn't a fixed bias to correct with one constant — it's strongest in the wind band where the fleet spends most of its downwind time (11–15kt) and fades out by the high teens. A correction to BTF's 152.5°/24% number needs to fall off with wind speed the same way this correlation does, not apply a flat discount everywhere.
Santa Barbara to King Harbor is sailed almost every year — seven runnings are logged (2016, 2018, 2021–2023, 2025–2026), each a short (~90–120nm, 8–10hr) overnight coastal race. No single year has enough deep-angle excursions to test alone (one year has only 4), so this section pools all seven — same course, same fleet, different years and wind — the way Section 05–07 treated each ocean crossing as its own test.
A clean, monotonic −27% falloff, closer to TP’15's steep decay than to SD2PV's gentler one — and the cross-correlation test on this course's cleanest window (SB2KH 2016, 2.06hr at a steady 13.9kt) lands right in the same near-zero cluster as TP’15/SD2PV on the dose-response chart above. Both independent tests agree: the same borrowed-momentum pattern is fully present here.
So: no, sustained deeper does not straightforwardly apply on SB2KH either. The heat-and-dive mechanism is there (92.3% of dives still preceded by a hotter angle), the timing signature matches the light-wind races, and the direct duration test shows the same kind of decay TP’15 shows. Any deep-angle boatspeed number pulled from this race's pooled data is likely inflated the same way TP’15's is.
There's a second, independent reason to be skeptical of a deep-angle strategy specifically on this course, unrelated to the Hookup Maneuver: the earlier BTF direct-vs-jibe simulation found SB2KH’21 and SB2KH’25 went 100% direct — the rhumb line to King Harbor already runs close to the wind axis most years, so there's little routing benefit to chasing a deeper angle even before accounting for whether the speed at that angle is real. On this course, both the tactical case and the data-quality case point the same way.
Newport to Ensenada is the fleet's other regular fixture: four runnings logged (2017, 2018, 2021, 2026), a ~125–160nm overnight coastal race, wind fairly consistent year to year (10.6–12.7kt mean per year — the narrowest year-to-year spread of any race family tested). 2021 barely registers (2 deep excursions total), so this is really 2017/2018/2026 pooled, pulling the same weight as SB2KH's seven years.
Only three buckets here — every excursion that lasted past 60s in the whole N2E dataset happened to occur below 8kt (light air, dropped by the band restriction), so there's no in-band long-duration bucket to show. What's left is unambiguous: 8.41→6.14kt, −27%, matching SB2KH almost exactly. The whole-race unrestricted version (not shown, no band filter) reads as an even steeper −50% — but that number is partly a wind-confound artifact running the opposite direction from SD2PV/SB2KH's: N2E's longest-held excursions happen to cluster in lighter air, so some of that apparent extra decay is really just lower wind, not duration. The band-restricted −27% is the trustworthy comparison figure.
N2E shows the strongest version of the mechanism of any race yet (10.8° average heat-up, the biggest of the six) and its concurrent correlation (r=−0.10) is the most negative of the whole set — further from TP’17's strong-wind regime than even TP’15 itself. Combined with SD2PV and SB2KH, that's now three independent light-wind races (plus TP’15) all clustered at or below zero, against Coastal Cup and TP’17 clearly above it at higher wind. And as with SB2KH, N2E 2018 was already flagged in the original BTF what-if simulation as a course sailed 100% direct — another race where the tactical case against chasing a deep angle and the data-quality case against trusting its measured speed reinforce each other rather than pulling in different directions.
Everything so far diagnoses a measurement bias: the pooled polar overstates achievable Bsp/VMG at deep angles because it's contaminated by momentary bursts. It doesn't yet say whether the maneuver is a good idea. That's a different, harder question, and it's the one that actually matters at the helm: across a real leg of sailing, does cycling hot→deep→hot beat just finding one honest, sustainable angle and holding it?
Here's the test: restrict to TWA 120–165° — the zone where a boat is genuinely trying to sail downwind-optimized, not reaching for some other tactical reason — and compare two numbers for that same zone, in the same race: (A) the actual time-averaged wind-axis VMG the boat achieved, real behavior included, whatever mix of heat-diving and holding actually happened; against (B) the best achievable VMG from a single 15°-wide angle band, using only samples that had already dwelled ≥60 continuous seconds in that band — i.e. the honest, sustainable ceiling for “just pick an angle and hold it.”
Every single bar is negative. There is no race in this dataset where the real, actually-sailed mix of technique beat the honest sustainable-angle benchmark. The three long ocean passages — N2E, TP’17, TP’15 — cluster tightly at a modest −2 to −4%. SD2PV sits a bit further out at −6.8%. Coastal Cup and SB2KH show much larger gaps (−18% and −25%) — but both are the courses already flagged elsewhere in this report as often sailed 100% direct, so a chunk of their “shortfall” is likely rational VMG-to-finish routing (deliberately not chasing the wind-axis-optimal angle because the rhumb line doesn't call for it), not evidence about the maneuver itself.
The answer: no, heat-and-dive does not beat sustained deep — and on the three races where this is a clean test, real sailing falls a little short of it. That's the honest reading of every angle this report has taken. The maneuver produces genuinely higher instantaneous boatspeed in the moment (Section 1–3 showed that directly), and the helm habit is universal (92–98% of every race tested), but the extra speed during the dive doesn't outrun what's given up heating up and decaying afterward. A crew that identified the honestly-sustainable 135–150° zone and disciplined itself to just hold it, rather than surfing between hot and deep, would likely match or slightly beat what actually happened — by the numbers here, on the order of 2–4% more wind-axis VMG on a long passage.
Two honest limits on that claim, both already established earlier in this report: this is wind-axis VMG, not VMG-to-finish — if the wind isn't blowing straight at the destination (usually isn't), the tactically correct angle can differ from the wind-axis optimum, which is most of what's inflating SB2KH/Coastal Cup's numbers above. And the “best sustainable” benchmark is itself built from the relatively few times a boat happened to hold ≥60s in that band — a small, possibly non-representative sample of conditions, the same selection-bias caveat [[project_btf]] flagged for its own deep-angle data. Neither limit changes the sign of the result across three independent long-passage races, though — just the precision of the exact percentage.