Prior beliefs overturned or newly pinned by the 2026-07-02 synthesis — all verified against live Gamma/CLOB/obs pulls (12:20–14:00Z). The verdicts that follow from these: weather-strategy-verdicts-2026-07-02. Detail: /Users/levander/coding/pmv2/research/weather-strategy-synthesis-2026-07-02.md.

1. Liquidity-rewards pools on weather EXIST

Prior belief "no LR on weather" was WRONG

467 temp markets carry clobRewards with rewardsDailyRate > 0: median 31k/day total**.

  • Top city pools/day: seoul-lowest 1,010, LA 828, amsterdam 719, madrid $703.
  • Params: min 20–50 shares, max spread 4.5c. Seoul highest-temp: no pool.
  • 25% maker rebate applies (weather is fee-bearing, feeType=weather_fees).
  • This is what makes the EU maker lane a real opportunity ( LR farming — PILOT after infra (the actual EU opportunity)).

2. Market universe ~7× bigger than backtested

  • 48 cities with a Jul-2 highest-temp market (156 daily temp events; 11 buckets each) — 41 new vs the backtested set: US+8 (LA/SF/dallas/austin/houston/denver/seattle/toronto), EU+9 (munich/milan/madrid/warsaw/amsterdam/helsinki/moscow/istanbul/ankara), AS+15, other+9.
  • Plus a lowest-temperature market series (8 cities) — overnight-min analog, morning lock, unexplored as a market surface.

Reconcile with the 2026-06-19 lowest-play NO-GO

The US lowest taker play was NO-GO’d (post-lock evening falls, nyc+miami backtest — see weather-bet-live-status). What’s new here is the series’ breadth (8 cities) and that seoul-lowest carries the single biggest LR pool ($1,200/day) — relevant for maker quoting, not for reviving the taker play.

3. Resolution stations pinned (engine-critical)

  • paris = LFPB Le Bourget (NOT LFPG/CDG)
  • london = EGLC London City (NOT Heathrow)
  • nyc = KLGA; denver = Buckley SFB; seoul = RKSI Incheon
  • Istanbul / Moscow / Tel-Aviv resolve via NOAA weather.gov timeseries, not Wunderground
  • Hong Kong resolves via HKO at 0.1°C precision (unique; research-only testbed)

Consistent with the station table in weather-strategy-validated-plan — now live-verified per city and binding for any EU/AS engine.

4. fee_usd=0 ledger bug overstates live ROI

  • place_order in position_manager hard-codes fee_usd = 0 (lib.rs ~L521) → live weather ROI overstated ~0.45–0.67%/side.
  • True fee = 0.05·p·(1−p) of notional (~0.64% at p=0.85).
  • Build #1: fix before any pilot scoring — the US pilot’s pass gate is meaningless without it.

5. api.weather.com public web key polls the WU resolution truth directly

  • ~30-min cadence, every city — eliminates WU-vs-METAR mismatch risk in the truth feed.
  • ToS-gray; keep the METAR fallback + obs-staleness kill-switch.
  • Complements fix-resolve-twc-primary-source: TWC/weather.com is what Polymarket actually resolves on; this is a direct poll of that source.

6. °C boundary-gate degeneracy fixable for 7/8 EU cities

Same-station public tenths feeds verified live:

FeedStationsCadence / note
DWD 10-minEDDM (munich)10-min
FMIEFHK (helsinki)open API
IMGWEPWA (warsaw)open API
JMA AMeDASRJTT (tokyo, station 44166)10-min, ~3–10-min lag, validated to 0.1°C
SYNOP via OGIMETLFPB, EHAM, LIMC, LEMDhourly
  • london: no SYNOP — Met Office DataHub is the unverified path. seoul: needs a KMA APIhub key.
  • US already has METAR T-group tenths (KLGA T03060217 = 30.6°C).
  • Caveat: on EU locks the tenths gate recovers only +1–1.8pp — lock HOUR is the lever, not the gate; this fix alone did not save the EU taker lane.

7. Tick/bucket structure

  • US: 2°F buckets, 0.01 tick; all non-US: 1°C buckets, 0.001 tick. negRisk=true everywhere.