Kalshi Weather Agent

Live
Total P&L
$0.00
No trades yet
Win Rate
0W - 0L - 0 pending
Active Positions
0
Next settlement: 9:00 AM CT
Return on Investment
$0 invested
How the Bot Works
Ensemble forecasting + split timing strategy
ECMWF IFS 51 members GFS 31 members ICON 40 members 122 Forecasts P(YES) Calculation

Ensemble Model

  • 122 independent weather forecasts from 3 global models
  • Empirical probability distribution (not just one forecast)
  • Accounts for model uncertainty and forecast spread
  • Fallback to calibrated standard deviation if ensemble unavailable

Split Timing Strategy

  • 8:00 PM CT: Bet on LOW temps for tomorrow (9hrs before overnight low)
  • 9:30 AM CT: Bet on HIGH temps for today (2-4hrs before afternoon peak)
  • Reduces exposure time and captures freshest forecasts
  • $4 budget per run ($8/day total)

Probability Math

  • Normal distribution centered on NWS forecast
  • Forecast error SDs calibrated by city and season
  • Example: Chicago winter = 7°F SD, summer = 4°F SD
  • Ensemble overrides SD model when 10+ members available

Risk Guardrails

  • EV threshold: Only bet when expected value > +5¢
  • Price range: 15-85¢ (avoid extreme odds)
  • City limits: Max 2 bets per city per day
  • Dedup: Never bet opposite sides or duplicate tickers
P&L Over Time
Win Rate Trend
All Trades
Run History
Calibration
Date City Ticker Side Cost Prob Source EV Result P&L
Time Mode Target Date Cities Placed Skipped Cost
Date City Type Forecast Observed Error Ensemble