A research tool for tuning NHC1 progressive-handicap blending parameters across a sailing season.
If we use a more aggressive set of NHC blend parameters in the first series of the season, the resulting handicaps at the end of the season should form a better basis for "fair racing" for the rest of the season using less aggressive parameters. The idea is that at the start of the season the handicaps will initially be way off "fair", but with aggressive blending we can get to fair handicaps by the end of the first series.
Given four Sailwave JSON exports — one per series of a four-series HYC season — blender searches, per fleet, for the NHC1 parameter set that, applied to series 1 only, produces the fairest racing in series 2–4 (which are scored with standard NHC1 defaults).
"Fairest" is operationalised as the lowest within-series coefficient of variation of gross series points across a stable cohort of competitors. See docs/design/objective.md for the precise definition.
For each fleet, the tool reports:
- A recommended series-1 parameter set, with the objective value and a comparison to the SWNHC2015 defaults.
- 1D parameter-landscape slices through the optimum.
- Per-boat TCF trajectories under standard vs. optimised series-1 parameters, plotted across all four series.
- Bootstrap-based confidence on the optimum.
Design phase — no code yet. The problem statement, objective function, evaluation pipeline, and optimisation scope are documented under docs/design/. Implementation begins once the design is settled and we have the real Sailwave files in hand.
- Recommended Series-1 profile — best honest answer from the data we have so far
- 2025 HYC Tuesday Puppeteer HPH — first real-data finding
- 2024 HYC Tuesday Puppeteer HPH — second season, opposite optimum
- 2024 vs 2025 comparison — cross-season analysis with bootstrap CIs
- Problem statement — the thesis, the fairness intuition, what "done" looks like
- Objective function — the scalar the optimiser minimises
- Evaluation pipeline — how one parameter set is evaluated end-to-end
- Optimisation — search space, constraints, deliverables
- NHC1 algorithm — algorithm reference
blender is a sister project to Sail Scoring, the main sail-racing scoring application. Sail Scoring runs NHC1 in production; blender re-implements the same algorithm for offline parameter optimisation. The tool is research-only and is not integrated into the app.
MIT — see LICENSE. Copyright Mark McLoughlin.