Signal Methodology v1.0

When an independent multi-model ensemble and the HRRR hourly update agree that a temperature threshold will NOT be reached, and the market is still pricing that event at 3-18¢ (implying 3-18% probability), the true probability is materially lower, creating positive EV on the NO side.

Source: Four independent global numerical weather models - GFS (NOAA), ECMWF, ICON (DWD) and GEM (Environment Canada).

Resolution: ~9-25 km horizontal; each model re-runs every 6-12 hours.

Lead time used: 12-96 hours (Day 0 through Day 4)

Role in signal: Primary temperature forecast. Pooling four independent models cancels out any single model's bias; the spread across them quantifies uncertainty. Narrow spread (low σ) = strong agreement = high confidence.

Key metric: Ensemble mean 2m temperature vs. the market threshold. A calibrated minimum separation (the "buffer") is required before a signal is scored - exact thresholds are proprietary.

Source: NOAA/ESRL HRRR - single deterministic model

Resolution: 3 km horizontal, hourly runs

Lead time used: 0-36 hours (same-day and next-day markets only)

Role in signal: Confirmation for near-term signals. HRRR assimilates real-time radar, surface observations, and satellite data - it has materially higher skill at ≤24 hours than the global models.

Key metric: HRRR maximum 2m temperature vs. the market threshold. Agreement with the ensemble required for signal publication.

Fetch open temperature markets from Kalshi (KXHIGH*/KXLOW* series via the Kalshi API). The same contracts are mirrored on Coinbase and Robinhood, so a Kalshi signal is tradeable on any of the three. Filter for markets resolving within the next 96 hours with adequate liquidity.

For each market, extract the forecast high temperature for the relevant city, date, and hour from the multi-model ensemble mean (GFS, ECMWF, ICON, GEM - all markets) and HRRR (markets resolving ≤36h).

Compute the temperature buffer: the gap between the forecast high and the Kalshi threshold:

A positive buffer means the model forecasts the threshold will NOT be reached. Larger buffer = higher confidence = higher win-probability.

For ≤36h markets, HRRR must independently agree with the ensemble. For Day 2-4 markets, the model ensemble spread must fall within a calibrated tolerance; high spread = uncertain = no signal.

Compute a calibrated win-probability, check orderbook depth for minimum fill ≥$8, and publish signal via Telegram with full reasoning: city, threshold, model mean, HRRR (if applicable), buffer, win-probability, and suggested position sizing.

All tiers manage a single bankroll - no slots. We hold no more than 2 positions in any one weather region, so correlated weather can't take out several at once.

Subscribers manage a single bankroll and stake a fixed fraction of it on each signal - no slots, no progressive ladders.

When the bankroll grows, the same fraction naturally stakes a little more; when it shrinks, you stake less. A loss never increases the next bet - which removes the temptation to "chase" and caps the damage of any losing streak.

The stop sits a fixed 20¢ below entry. When an active position's NO price falls 20¢ under your fill (e.g., 95¢ → 75¢), the signal system flags an immediate exit. This caps the loss at roughly 20¢ per contract - about 21% of a 95¢ position - regardless of how much time is left.

Signals where the Kalshi orderbook cannot fill ≥$8 (thin markets, usually <10 contracts at the target NO price) are flagged as SKIP. These are published for informational purposes but are excluded from the W/L track record.

Kalshi charges a flat $0.01 per contract fee on fills. This fee applies to both entry and exit (stop-loss) fills.

At high NO prices (95¢+), the fee is material (25-50% of gross profit). Always check net-of-fees EV before sizing up.

Minimum fill $8. If the orderbook cannot fill $8 at the target price, skip the trade entirely. Thin markets mean: poor price execution, high slippage on exit, and outsized fee impact.

The stop - 20¢ below entry - is not a suggestion. When the NO price falls that far (e.g., 95¢ → 75¢), the position is exiting a scenario where our model edge has deteriorated (an unexpected weather development). Holding through a stop risks full position loss.

If a position moves against us, we do not add contracts. The stop-loss exits the position. We wait for the next independent signal.

Every position is a small, fixed fraction of one bankroll (commonly 1-5%). A loss never increases the next stake, and no single trade can lose more than ~21% of its own position. We also cap concurrent positions in any one weather region at 2, so a single correlated weather event can't hit several at once.

3rd Eyes publishes impersonal market research and signals only. All signals are identical for every subscriber - no advice is tailored to any individual's financial situation, account size, or risk tolerance. 3rd Eyes does not hold discretionary trading authority over any subscriber's account, does not execute trades on behalf of subscribers, and does not act as a Commodity Trading Advisor (CTA) as defined under the Commodity Exchange Act. 3rd Eyes operates under the publisher exemption from CTA registration pursuant to CFTC Rule 4.14(a)(9).

Trading Kalshi event contracts involves substantial risk of loss, including the possibility of total loss of all capital deployed. Past performance - published and exchange-verified at /track - is not indicative of future results. Losses do occur; the stop-loss system is designed to limit, not eliminate, losses.

Hypothetical and projected performance figures (e.g., bankroll calculator outputs) are for educational illustration only and do not represent actual or guaranteed trading results.

Signals are for US-legal, CFTC-regulated weather contracts: Kalshi, and the identical contracts carried by Coinbase and Robinhood. Subscribers execute on whichever of these they hold an account with, and verify availability and compliance in their own jurisdiction. 3rd Eyes signals do not constitute legal, financial, or regulatory advice.