Natural Gas Weather Play

Primary free source: NOAA Climate Prediction Center (cpc.ncep.noaa.gov) provides 6-10 and 8-14 day US temperature outlooks - the demand side that sets Henry Hub. For the Canadian source region, use Environment and Climate Change Canada (weather.gc.ca) and The Weather Network, which track the building Arctic highs and polar-vortex lobes. Weather models (GFS, ECMWF, and the Canadian CMC/GDPS) can be viewed on sites like tropicaltidbits.com or weathermodels.com. For serious trading, consider paid services like Commodity Weather Group or Atmospheric G2.

Enter when the forecast first shifts significantly (typically 6-10 days before the event). The 6-10 day forecast is the sweet spot - accurate enough to be reliable, far enough out to capture most of the move. Don't wait until 1-2 days before, as the move has already happened. The Canadian edge is that you can sometimes spot the Arctic high building over the Prairies even earlier than the US outlook turns cold.

Canada has no domestic retail natural gas futures, so you trade the global Henry Hub benchmark either through NYMEX futures (NG full size, QG E-mini, or MNG Micro) via a CIRO-registered futures dealer in USD, or through TSX-listed ETFs in CAD (HNU is +2x daily bull, HND is -2x daily bear, HUN is ~1x). All of these settle to or track Henry Hub in Louisiana, which is priced by US supply and demand - the world's largest gas market. So even though you're in Canada, your price follows US (and increasingly LNG-export) demand. AECO is Canada's own benchmark but is essentially an institutional/physical market.

Generally, look for at least a 15% deviation from normal HDDs. Smaller deviations (5-10%) are often noise and have lower win rates. Deviations of 20%+ are stronger signals. Build your own thresholds based on tracking your trades.

Summer weather trades can work but have lower impact than winter. Focus on major heat domes (CDDs 20%+ above normal) in power-heavy regions like Texas and the US Southeast (secondarily Ontario and the Prairies). Summer positions should be smaller than winter positions due to lower price sensitivity.

Look at temperature forecasts for key US demand regions (Northeast, Midwest) on all three models. If they show similar cold/warm patterns, it's high confidence. If they disagree significantly, wait for convergence. ECMWF is generally considered most accurate for extended range (7+ days); the Canadian CMC is a useful independent tie-breaker and often resolves the source-region cold clearly.

Low storage (<5% below the 5-year average) = full position or larger on cold forecasts (amplified impact). Normal storage = standard position. High storage (>5% above average) = reduced position on cold forecasts (dampened impact). The price-moving number is the EIA US storage report; AECO/Canadian storage colours the Western Canada picture but doesn't move Henry Hub directly. Always check storage context before sizing.

Exit triggers: (1) Weather event peaks (coldest/hottest day passes), (2) Forecast moderates significantly, (3) Price target reached, (4) Time stop (no move after 3-4 days). Don't hold hoping for more after the event passes - that's when prices retreat. If you're in HNU or HND, exit even more decisively: daily-reset decay penalizes a leveraged ETF held through any sideways stretch.

The EIA storage report (Thursday 10:30 AM ET) creates volatility in Henry Hub. Options: (1) Reduce position by 50% before the report, (2) Hold if your weather view is strong and the EIA is likely to confirm (cold weather = bullish draw), (3) Use a tight stop. Don't enter new weather positions right before the EIA print.

Regional cold (e.g. just the US Northeast) = moderate price move (8-12%). Continent-wide cold (Northeast + Midwest + South, fed by a massive Prairie high draining down the Plains) = major price move (15-25%+). Multi-regional events are rarer but powerful, and a severe Prairie/Montney freeze can clip WCSB supply on top. Size positions larger for continent-wide events.

Collect data: weekly HDD deviation from normal and corresponding Henry Hub price change for 2+ years. Run regression: Price_Change = alpha + beta x HDD_Deviation. Coefficient beta shows price sensitivity. Segment by season (winter beta higher than summer). Update quarterly with new data. If you trade HNU/HND, model the ETF's realized (path-dependent) return separately - it is not a clean 2x of Henry Hub over multi-day events.

Key patterns: (1) Polar vortex strength - a weak vortex lets lobes detach over Canada and spill south. (2) SSW events - precede cold 2-4 weeks later. (3) NAO index - negative = Eastern Canada/US East cold. (4) Pacific jet / ENSO (El Nino/La Nina) - decides whether Arctic air loads the Prairies and Plains or gets deflected. Understanding these provides early signals before standard forecasts, and the source region (Canada) is where they show up first.

Use options (NYMEX Henry Hub options via a futures dealer) when: (1) Forecast uncertain but a big move expected (straddle), (2) You want defined risk on an uncertain forecast, (3) A large position needs protection (protective put), (4) You want leverage with limited capital. Options cost premium but eliminate gap risk and define maximum loss. Remember each NG option covers 10,000 MMBtu, so premiums are large; a Canadian wanting capped risk without options can instead cap size in HNU/HND.

Limit per-event risk to 3%, total weather exposure to 8%. Manage correlation - back-to-back cold events are correlated, don't double up. Diversify across timeframes (near-term + medium-term). Roll from completed events to new setups. Maintain a capital buffer for opportunities. If positions mix USD futures and CAD ETFs, convert everything to one base currency (CAD) so the caps are real.

Track every trade in a database with forecast details and outcomes. After 50+ trades, analyze: win rate by HDD deviation, model agreement, storage context, season. Build a personal scoring model. Edge comes from superior interpretation, reading the Canadian source region early, faster action, and disciplined execution - not from unique data, since everyone sees the same NOAA output.