S/V Chim Chim · Race Fleet vs. Morrelli & Melvin VPP
Thirty-six races have been scored against the Morrelli & Melvin velocity-prediction model that defines what this catamaran should do in a given wind. Only one of them came out ahead of its own model — and it isn't close.
TransPac ’15's overall score — the only race above 100% of target in the fleet
Fleet average: 88.0% · next-best race: 96.2%
Each race's actual P85 boat speed, as a percentage of what the VPP predicts for its own wind and angle. Bars start at zero — nothing here is a truncated axis.
TransPac is a downwind delivery to Hawaii — 147,575 of TP15's 147,735 scored points (99.9%) fall in the downwind category, and none fall in upwind at all. Every other race in the fleet has to blend in a weaker upwind or reach number that drags its overall score down. TP15 never pays that toll: its overall score is, for practical purposes, its downwind score.
That alone doesn't explain the 101.6% — it explains why nothing dilutes it. The fleet's own downwind average is 88.6%. TP15 sits 13 points above that, which is the number that actually needs explaining.
Every one of the 36 races, plotted by what share of its scored sailing was downwind vs. its overall VPP score. Each dot is one race; the dashed line is the fleet's best-fit trend.
A real, if noisy, relationship — going from an all-upwind/reach race to an all-downwind one predicts roughly a 16-point jump in score along the trend line. It holds even with TP15 removed entirely (r = 0.49), so this isn't just one outlier dragging the correlation around.
The sharper version of the story: share of upwind sailing is actually the stronger single predictor (r = −0.60), more than downwind share itself (r = 0.54) and far more than reach share, which barely correlates with the outcome at all (r = 0.02). Avoiding upwind sailing predicts a good score slightly better than sailing downwind does — TP15 happens to do both at once.
TransPac ’15 vs. the 36-race fleet average, per category. No upwind bar for TP15 — there's no upwind sailing to score.
A prior pass through this same fleet (comparing raw polars to the VPP curve directly) found a consistent pattern: actual boat speed underperforms the M&M target on a hot reach, and overperforms it once the boat gets deep enough — somewhere around a 117–148° crossover, depending on wind strength. TP15 spends nearly all of its downwind time on the far side of that line.
This isn't unique to TP15 — TransPac '17 sails a similarly deep average angle and still lands at 96.1% downwind, well short of TP15's 101.6%. Angle explains the direction of the gap, not its full size.
Sustained periods in the 12–15kt trade-wind band, TP15 vs. TransPac '17 — the closest comparable race by course and category, both ocean passages to windward destinations under trade-wind flow.
The P85 statistic behind every leaderboard score rewards a boat that can hold its speed, not just touch it once. Three hours of unbroken 12–15kt pressure gives the boat, the crew, and the trim time to settle into a true steady state — and P85 captures that ceiling. A race made of short gusty bursts never gets the chance to show the same number, even if its peak speed is identical.
Almost every one of those 36,306 samples is downwind — the histogram peaks at 130–135°, right where the crossover-angle finding above says this boat starts beating its own model. This is what "holding pressure" actually looked like: not a brief hero angle, but three-plus hours at a time, centered tightly on 134°.
A following sea seems like an obvious candidate — but this was already tested directly, on both TransPac races, by regressing measured boat-speed residual (actual minus polar-predicted) against a WW3 hindcast swell-assist index.
TP15 — the race that beat its own model — shows no detectable swell effect at all. TP17 has the one real, robust swell–speed relationship found anywhere in this fleet, and it runs the wrong way for "following seas pushed the boat": likelier a bow-burying penalty in quartering seas (these catamarans have no monohull-style buoyant bow flare to lift over a wave from behind), clustering right at TP17's typical ~135° broad-reach angle. TP17 finished at 96.1% downwind — solid, but nowhere near TP15's 101.6%.
So the fleet's one confirmed swell effect hurts a race rather than helping it, and it isn't present in TP15 at all. That's evidence against swell push as the explanation here, not for it.
The model itself may be slightly conservative. The Morrelli & Melvin VPP behind these targets was built around a 15.5-metric-ton boat. Chim Chim was actually weighed in 2020 at 14.87mT — about 4% lighter. A lighter boat plausibly sails a little faster than a heavier-boat model predicts, most visibly downwind. If that's a real effect, it nudges every race's ratio up a little, and TP15 simply has the largest, steadiest sample to show it clearly.
The leaderboard metric structurally favors downwind-only races. Any race that happens to skip upwind sailing entirely gets scored purely on its best category, with nothing to average it back down. This isn't a hunch about TP15 — it's a measurable, fleet-wide correlation (r = 0.54 with downwind share, r = −0.60 with upwind share, n = 36). That's an honest property of the metric, not an error — but it means "best on the leaderboard" and "best-sailed race" aren't quite the same question.
race_vpp_performance (36 isRacing=1 groups), built by scripts/rebuild_vpp_performance.php — P85 actual boat speed per 5° TWA / matched-TWS cell (min n=5), divided by the matched MM_polars target, n-weighted across cells. Category boundaries: upwind <80°, reach 80–100°, downwind >100°. Downwind/upwind/reach share per race computed as each category's n divided by the race's total scored n (upwind/reach null treated as 0); Pearson r against overall ratio computed across all 36 races. Crossover-angle, boat-weight, and swell-residual findings drawn from prior sessions' analysis of this same fleet (raw-polar-vs-VPP comparison; 2020 haul-out weighing; WW3-hindcast swell-assist vs. polar-residual regression on TP15/TP17) — not re-derived here.