Skip to content

Pre-Race Data Collection Workflow

Newport → Bermuda Race — Lupo Di Mare

Rev: v2.0  ·  Updated: 2026-05-29 Tactic codes: H-* hypotheses in Routing Hypothesis Labels; T-* tactics in the Routing Tactic Catalog.

The day-by-day procedure for collecting, validating, and analysing pre-race data from T-10 through start morning.

Rule constraint (2026 NOR, RRS 41 modification): downloading GRIB (GFS, ECMWF, RTOFS, SST) during the race and running onboard routing software against it is permitted. Not permitted: customised routing advice from shore, including AI-generated. All analysis in this project — the matcher, the daily brief, every W-* call — is PRE-RACE only.

The one-command spine

One script runs the whole forecast chain — anyone on the crew can run it daily to build reps:

.venv/bin/python scripts/daily_brief.py

It does fetch → classify → match → brief (see Forecast Analog Matcher) and writes a dated brief to data/daily-brief/. So the steps below that say "pull GFS/ECMWF/ICON, classify the pattern, find the analog" are automated — what stays manual is the judgement layer:

  • Gulf Stream imagery — RTOFS / SST / AVISO altimetry, interpreted via the Gulf Stream Framework (refreshed by scripts/gulfstream_monitor/).
  • Routing-software runs — H-* hypotheses in Expedition against the pre-race GRIB.
  • The decisions — the per-day outputs below.

Race Calendar

Milestone Date
ORR data submission deadline ✅ done May 29, 2026
T-10 June 9, 2026
T-7 June 12, 2026
T-5 June 14, 2026
T-3 June 16, 2026
T-1 June 18, 2026
Start (1300 EDT) Friday June 19, 2026

T-10 → T-7: Baseline

T-10 (June 9): 1. Run daily_brief.py — captures GFS/ECMWF/ICON + marine, classifies the pattern, ranks analogs. The brief's trend block tracks model agreement day-over-day (replaces the old hand-kept comparison log). 2. Pull OPC surface analysis for the synoptic pattern. 3. Pull RTOFS surface current + CoastWatch SST; record the north wall at 65°W. 4. Check AVISO/CMEMS altimetry for mesoscale features (CCR, WCE); note any found. 5. First routing passes in Expedition: H-RH (rhumb) and H-CL (conservative), GFS, S3 polar.

Read T-10 as: long-range signal only — identify analog candidates; do not commit to a strategy. No model reaches race day yet.

T-7: First meaningful routing runs

  1. Re-run the brief; compare to the T-10 baseline (the trend block flags what moved/intensified/weakened).
  2. Full routing sensitivity table (H-RH, H-PF, H-CF, H-CL × GFS, ECMWF); add maneuver-time cost to each.
  3. Validate RTOFS north wall against SST — agree within 30 nm? If not, flag it and trust SST.
  4. Identify mesoscale features and predicted tracks. Write the T-7 brief.

Decision output: most-likely analog (confidence) · GFS–ECMWF delta (<3 h / 3–6 h / >6 h) · which hypotheses remain viable · any single-model-dependent (fragile) elements.

T-5: Gulf Stream position lock

  1. Pull RTOFS, latest CoastWatch SST, AVISO altimetry.
  2. Run the position analysis from the Gulf Stream Framework: north wall at 65°W → CCR in the 36–39°N corridor → WCE in the 31–34°N corridor → wind-against-current risk.
  3. Cross-check the model current against imagery using RTOFS Interpretation.
  4. Update routing runs with the new Stream data. Write the T-5 brief — first one where Stream strategy is substantive.

Decision output: north wall °N ± ___ nm (conf) · core current ___ kt @ °T · CCR Y/N · WCE Y/N · preliminary Stream entry latitude.

T-3: Strategy narrowing

  1. Re-run the brief + routing sensitivity table; how did the ranking change?
  2. Confirm or revise the primary hypothesis (is H-CL still preferred, or has a cleaner picture made H-PF viable?).
  3. Update the maneuver-cost-adjusted comparison; re-validate RTOFS vs SST; check feature positions.
  4. Write the T-3 brief; begin assigning watch-level tactical cards.

Decision output: primary strategy locked (H-___, or undecided between two) · trigger conditions to change it · Stream entry latitude · estimated elapsed range (GFS–ECMWF, maneuver-adjusted).

T-1: Final confirmation

  1. Final GRIB cycles (GFS + ECMWF); final Stream position (RTOFS + SST); confirm CCR/WCE positions.
  2. Two final routing checks: primary hypothesis (GFS + ECMWF) and fallback H-RH (GFS).
  3. Write the T-1 brief — the document the crew reviews at the pre-race meeting.
  4. Print/download all onboard cards; load pre-race GRIB + SST + altimetry onto the nav tablet; label clearly.
  5. Brief crew: analog type, primary strategy, Stream entry plan, squall thresholds.

Decision output: FINAL strategy H- · FINAL Stream entry °N · FINAL Gulf-Stream-entry sail plan (per Card 04) · squall threshold (30 nm echo if not drilled, 20 nm if drilled) · estimated elapsed.

Race morning (T-0)

Final GFS 6z/12z cycle (note overnight change) · final OPC surface analysis (front timing) · confirm instruments recording · confirm pre-loaded GRIB/SST/altimetry on the nav tablet · confirm all 10 onboard cards aboard (physical preferred) · brief the start-area plan and first-12 h watch.

Rule-compliance final check: - [ ] All AI analysis was pre-race only — no AI routing query during the race - [ ] Pre-loaded GRIB is the last allowed weather for route optimisation - [ ] No post-start shore communication for routing advice (RRS 41) - [ ] NOR/SI reviewed

Daily data-pull quick reference

Source Record How
GFS + ECMWF + ICON Corridor wind/gust/MSLP/CAPE; model agreement Automated — daily_brief.py
Marine (SST / current / waves) Corridor sea-state; north-wall SST jump Automated — daily_brief.py
OPC surface analysis Front / high / low positions Manual, T-7→T-0 daily
RTOFS surface current North wall at 65°W; core current Manual / gulfstream_monitor, T-7→T-0
CoastWatch SST Thermal north wall; CCR/WCE signatures Manual / gulfstream_monitor, T-7→T-0
AVISO/CMEMS altimetry SSH anomaly; CCR/WCE confirmation Manual, T-5→T-0 every 2 days

Storage: the brief auto-writes data/daily-brief/YYYY-MM-DD.md; store routing runs under 08_routing_runs/.

GRIB pre-load checklist (race morning)

Product Loaded? Date
GFS GRIB 0–120 h ☐
ECMWF GRIB 0–96 h ☐
RTOFS surface current ☐
SST overlay (CoastWatch) ☐
CAPE / convective layer ☐
Altimetry overlay (AVISO) ☐

Pre-race + pre-loaded → ✓ compliant. Downloaded after the start gun → ✗ not compliant.

Sources

  • Pipeline — scripts/daily_brief.py, scripts/forecast_monitor/, scripts/gulfstream_monitor/; method in Forecast Analog Matcher.
  • Stream analysis procedure — Gulf Stream Framework.
  • Rule basis — 2026 NOR / RRS 41 modification (verify against the posted NOR/SI before the race).

PRE-RACE RESEARCH — not race-period routing advice. Updated 2026-05-29.