DBSC HalSail results → Sail Scoring

Captures and reconstructs Dublin Bay Sailing Club summer-series results, year by year, from the club's published HalSail pages into importable .sailscoring files — and uses them to populate a DBSC workspace in Sail Scoring.

It is the DBSC counterpart of iodai-archive: same shape of work (reconstruct published race history into .sailscoring), same payoff.

📋 CLARIFICATIONS.md — open questions for DBSC about how their results were modelled (things we can't resolve from the data alone). Add to it whenever a divergence comes down to a DBSC/HalSail decision, not our bug.

Why

The payoff is the cross-series identity and ranking work on the Sail Scoring horizon (docs/design/horizon.md in the app repo — the competitor-identity spine, the workspace season ladder, and the per-competitor multi-year career-arc page). Those features only come alive with years of real history in one workspace. DBSC's multi-season summer series — dozens of fleets, the same boats recurring across days and years — is exactly that fuel, and a far richer, messier handicap dataset than IODAI's one-design Optimists.

It also keeps a bulky, scraped corpus (~130–150 MB of raw HTML) out of the app repo, where reference/data/ is for small illustrative fixtures.

What's here

Two pipelines share the repo: the archive (2022–2025, frozen) and the 2026 live-parity loop.

sources/
  manifest.json         the four archived datasets (2022–2025)
  <year>/               archive capture (archive.halsail.com)
    catalog.json          normalized class→series join (the deliverable)
    _classdropdown.html   raw class list
    series/class-*.html   raw per-class series dropdowns (the join, raw)
    results/series-*.html raw results tables, one per (class × series)
  2026-live/            live-site (halsail.com) parity dataset
    *.sailscoring         generated day/group series files
    halsail/              captured live result fragments
lib/halsail/           HalSail HTML parser + DBSC day-series builder
scripts/
  halsail-archive-fetch.ts   archive capture (`pnpm capture`)
  halsail-fetch.ts           refresh 2026 live fragments (`pnpm fetch`)
  halsail-to-sailscoring.ts  fragments → .sailscoring (`pnpm to-sailscoring`)
  halsail-compare.ts         parity vs published (`pnpm compare`)
tests/                 vitest for the parser + converter

The conversion/parity tooling reuses the app's scoring engine by relative import (../sailscoring/lib/scoring), so it needs the sibling ../sailscoring checkout. The generic HalSail FAQ and the parity design doc stay in the app repo (reference/HalSail FAQ.pdf, docs/design/dbsc-parity-plan.md).

The HalSail archive model (what we learned)

The archive (archive.halsail.com) is a separate app from the live site, with a four-level AJAX cascade. Crucially, which tandem series a class is scored under is only exposed by a per-class AJAX call (_CrsSeryDropDown) and is absent from any results page — so we persist that class→series join both raw and as catalog.json. See SOURCES.md for the endpoint family.

The join settles how DBSC models its season (this is HalSail's tandem series — "ghost" series that re-score the same finishes under a different class, race subset, or discard rule, mapped in Sail Scoring onto fleets × sub-series):

• One "Overall" per class per day — e.g. Cruisers 0 IRC has a single Thursday Overall and a single Saturday Overall.
• Fixed vs progressive split — fixed-handicap classes (IRC/VPRS) span several days; progressive ECHO classes are per-day (their chain restarts each day).
• Named race-subset tandems — Thursday/Saturday Series A & B, Tuesday Series A/B/C, the whole-event Summer Series.
• Cross-day pools exist only when explicitly named — Thursday & Saturday Combined, RAYC Super League 1/2. Never by collapsing a finish-sheet boundary; model such a table as its own named sub-series.

Commands

pnpm install

# Archive (2022–2025, archive.halsail.com)
pnpm capture --map-only     # Stage A: datasets + classes + class→series join
pnpm capture                # Stage B: also every results table (~983 files)
pnpm capture --year=2025    # restrict to one dataset

# 2026 live parity (halsail.com)
pnpm fetch                  # refresh the live result fragments
pnpm to-sailscoring [day]   # fragments → .sailscoring
pnpm compare [day]          # re-score with the app engine, diff vs published
pnpm test                   # vitest (parser + converter)

Captures are single-threaded, 0.75 s between requests, resumable (a file already on disk is reused, never re-fetched). Read-only against public servers.

Status

Year	Catalog (class→series join)	Results tables	Reconstructed .sailscoring
2022	✅	✅	◻️
2023	✅	✅	◻️
2024	✅	✅	◻️
2025	✅	✅	◻️

Reconstruction into .sailscoring (and parity validation against the published tables) is future work; the captures above are the evidence base.

Licensing

This repository contains three kinds of material, licensed separately:

• Code — scripts/: MIT.
• Normalized data & docs — sources/**/catalog.json, sources/manifest.json, README.md, SOURCES.md: CC0 1.0 (public-domain dedication). These are extractions of published facts (series names, keys, structure), and facts are not copyrightable.
• Source pages — the verbatim *.html under sources/: not covered by either license. These are HalSail result pages published by DBSC, included only for reproducibility. All rights remain with their owners.