The 9 PM Polymarket Weather Pricing Window

Why Polymarket weather markets turn inefficient around the 9 PM UTC resolution window, and how slow forecast updates open the trading edge.

📅 2026-05-02 ⏱ 5 min read ✍ Delta T.
The 9 PM Polymarket Weather Pricing Window
← Back to blog

If you had to pick a single 30-minute window during the day where Polymarket weather markets are most inefficient, it would be roughly 19:30–21:00 UTC — what European traders call "the 9 PM window". This isn't a guess. It's measurable in the order book.

Why this window exists

The ECMWF release timeline

The 12 UTC run of ECMWF — the most weighted model in the bot's ensemble — becomes publicly available around 19 UTC (some downstream feeds, ~30 minutes earlier). For European weather markets resolving the next day, this run dramatically reduces uncertainty: temperature predictions for the next 24 hours tighten by 30–50% on the spread.

Why retail misses it

But Polymarket order books are dominated by retail traders who: (a) don't subscribe to ECMWF feeds, (b) are watching their forecasts on phone apps that lag the model run by 1–2 hours, (c) frequently aren't paying attention at 9 PM at all.

The measurable mispricing

Result: between roughly 19:30 and 21:00 UTC, you can routinely find European weather markets where the order book midprice is 4–8% off the new ECMWF-implied probability. The first market to reprice tends to do so when a single arbitrageur takes the spread; the rest follow over the next hour.

What the bot does

The 5-step playbook

Between 19:00 and 21:30 UTC the bot runs in elevated-priority mode for European markets. It:

  1. Polls ECMWF's first available downstream feed at 19:00 UTC
  2. Computes the implied probability for every active European weather market
  3. Compares against current Polymarket bid/ask
  4. If spread > threshold (default 6%) and the new forecast is meaningfully different from the previous run, it places a limit order at a slight discount to mid
  5. Waits up to 90 minutes for fill, then cancels if unfilled

Typical fill rates

Typical fill rate during this window: 60–75% of submissions get filled within 45 minutes.

Other windows worth watching

One warning

The 9 PM window is becoming more crowded as more bots learn about it. Two years ago typical edge was 8–12%; today it's 4–7%. Don't expect this to scale forever. The strategy still works, but the per-trade margin has compressed. Volume helps. Diversification across cities helps more.