Resource Center

Understand the markets,
the models, and the risk.

Plain-English explainers of how prediction markets work, where our edge comes from, how to read a signal, and how to trade responsibly. New here? Start with the basics below.

Prediction markets Why weather How we forecast Reading a signal Responsible trading Glossary
In-depth guides
→ Prediction markets, explained simply → Why we trade the weather (and only NO) → How to trade without blowing up

What are prediction markets?

A prediction market is a regulated exchange where you trade simple contracts on real-world outcomes - for example, "the high temperature in New York stays below 75°F today." Each contract settles at $1 if it turns out correct and $0 if not, so the price along the way (say, 95¢) reflects the market's implied probability of that outcome (about 95%).

Every contract has two sides: YES (the event happens) and NO (it does not). You can take either side. Kalshi is a CFTC-regulated US exchange - the same federal regulator that oversees the major futures markets - and it lists the temperature contracts our signals focus on. The identical contracts are also carried by Coinbase and Robinhood.

Why temperature markets?

Daily high- and low-temperature contracts have three useful properties: they are reasonably liquid, they settle quickly (usually the same or the next day), and they are frequently mispriced. Casual traders tend to chase extreme-temperature longshots, which leaves the NO side of those longshots priced too cheaply.

That is the gap we trade. When professional weather models are confident a temperature threshold will not be reached, but the market still prices that outcome as a real possibility, buying NO carries positive expected value. We focus on NO because the edge is larger and the forecast is easier to get right.

How we forecast

We run an ensemble of four independent global weather models (GFS, ECMWF, ICON, GEM) plus the high-resolution, hourly-updating HRRR model. The spread across those models is our measure of uncertainty: when they all cluster far from a market's threshold, confidence is high; when they disagree, we stay out.

A signal only fires when several checks clear at once: a sufficient temperature buffer between the forecast and the threshold, agreement across all models, HRRR confirmation, a low ensemble spread, and a pricing edge over the market. If any check fails, we do not trade - which is why a typical day produces only a handful of signals across 40+ cities. The full method is in the methodology.

Reading a signal

Each alert gives you everything you need in one message: the city, the exact contract, the model's win-probability, the NO price to pay, and a confidence grade. To act on it, open that contract in your trading app, buy NO at or below the listed price, and you are done - most markets settle the same or the next day.

New to placing trades? The quick-start guide walks through it step by step, and the sizing & risk page helps you choose how much to stake.

Responsible trading

These are real markets with real risk. You can lose your entire stake on any single trade. A high win rate is not a guarantee - even a long, honest record will have losing streaks, and small-edge NO positions can still go to zero on a rare-weather day.

Only ever trade with discretionary funds you can afford to lose entirely. Size positions conservatively, use the built-in stop-loss, and treat the calculator as a sensitivity tool - not a forecast. Past performance does not guarantee future results. 3rd Eyes publishes research and signals only; it is not personalized investment advice.

Glossary

Prediction market
A regulated exchange for trading contracts on real-world outcomes. Each contract pays $1 if correct, $0 if not.
Kalshi
A CFTC-regulated US exchange that lists the temperature contracts our signals trade. Coinbase and Robinhood carry the same contracts.
Contract
One tradeable position on a specific outcome, e.g. "Chicago high stays below 62°F today."
NO side / YES side
NO wins if the stated event does not happen; YES wins if it does. We trade NO on thresholds we are confident will not be reached.
Win-probability
Our model's calibrated estimate of the chance a contract settles in your favor.
Edge
The gap between our model's win-probability and the market's implied price. A positive edge means the market is mispricing the outcome.
Buffer
How far the forecast sits from the market's threshold, in °F. A bigger buffer means a safer trade.
Ensemble
Running several independent weather models and measuring how closely they agree.
HRRR
A high-resolution, hourly-updating US weather model used to confirm the ensemble forecast.
Settlement
When a market closes and pays out based on the officially recorded temperature.
Stop-loss
An automatic exit if a position moves against you by a set margin (20¢ below entry), to cap the loss.
Bankroll
The total pool you've set aside for these trades - only cash you can afford to lose. You size every position as a small fraction of it.
Position size
How much you stake on a single signal. Disciplined sizing is a fixed 1-5% of your bankroll per trade, scaled by the signal's confidence grade.
Start here → See the track record