Live Hurricane Prediction Markets at ForecastEx offer a new level of granularity for information discovery and risk transfer in the immediate lead-up to a storm’s landfall.

Background

Hurricanes are among the most socially and financially consequential of all weather phenomena, accounting for nearly 40% of insured losses from natural hazards (SwissRe). 

Last year, ForecastEx introduced Forecast Contracts on seasonal hurricane activity, including basin-wide storm counts and regional major (Category 3+) hurricane landfall contracts at various levels of granularity that resolve at the end of the year. 

The basin-wide storm count contracts serve the purpose of broad information (how does overall hurricane risk this year compare to other years?), while the major hurricane landfall contracts at the geographic level of the US county have substantial hypothetical value as risk-hedging and risk-transfer tools. 

On this risk hedging and risk transfer point, these major hurricane landfall contracts can still be improved on at least four dimensions

1. Increasing the number and reducing the size of geographic regions (reducing Basis Risk)
2. Choosing an intensity metric of severity that better represents structural damage and thus financial loss (reducing Basis Risk)
3. Increasing the number of wind speed thresholds to allow for better customization of desired financial positions (reducing Basis Risk)
4. Decreasing the time envelope from contract opening to closing (contracting liquidity and reducing time to payouts)

All four issues are addressed in ForecastEx’s just-released “Live Hurricane” contracts. 

These contracts are tied to individual named storms as they happen, resolving questions of the form:

“Will Hurricane Julia cause wind gusts of 100 mph or greater in Miami, FL?“

ForecastEx will list these contracts after a hurricane forms, receives a name, and begins to legitimately threaten predetermined coastal locations (see below). The contracts will trade continuously as forecasts evolve, and they will resolve definitely, not at the end of the year, but within a few days of the storm’s dissipation. 

• Access to live probabilities and the ability to trade is available via IBKR Prediction Markets.
• General primers on how these prediction markets work are available in the IBKR prediction markets FAQ (scroll down) and the ForecastEx FAQ.

How the contracts work

Each Live Hurricane contract is defined by three parameters: a named storm, a location, and a wind threshold. There is also a complementary event question that asks which location will experience the highest winds from a given named storm. 

Locations

The contracts reference 163 predefined locations representing recognizable coastal population centers or points of interest across the U.S. Gulf and Atlantic coasts, the Caribbean, Mexico, Central America, and northern South America (listed here).

Fig. 1 | Map of the live hurricane ForecastEx reference locations. These central locations and the size of the surrounding area are articulated in the location file published here. Figure produced by Patrick Brown with Matplotlib. 

Defining markets around a manageable number of predefined discrete locations is intended to pool participant interest and thus market volume and liquidity, facilitating trading.

Each location is centered at a particular latitude and longitude and has an area surrounding it where winds can count toward resolution for that location. The area around each central point constitutes a 10 km radius applied to a discrete 1 km × 1 km pixel grid including only land pixels.

Below are visualizations of some locations and the size of the footprint for which winds count towards resolution.

Fig. 2 | Settlement areas for two hurricane-prone metro areas in Florida: Tampa Bay (left) and Miami (right). Each location settles on the maximum 3-second gust across the 1 km × 1 km land cells within 10 km of its reference point. Figure produced by Patrick Brown with Reask data and plotted with Matplotlib.

How strong will the storm be?

For each location in play, contracts are listed on a ladder of wind-gust thresholds: 70 mph or greater, 80 mph or greater, and so on in 10-mph increments, up to 200 mph or greater.

As with all ForecastEx Forecast Contracts, the price at any moment maps directly onto the market’s implied probability that the threshold will be met or exceeded, so the full ladder gives a sense of not just whether a location will be hit, but also the full likelihood distribution of intensity.

Where will the storm hit hardest?

In addition to markets for the wind gust intensity magnitude at given locations above, ForecastEx will offer markets for which location, out of a defined pool, will eventually record the highest wind gust from the storm. This framing essentially drills down on the uncertainty in the storm’s path rather than uncertainty in the storm’s absolute wind intensity. 

Listing

Contracts for a location and threshold are only listed once forecast data indicates a meaningful probability (around 5%) that the location will experience 70 mph or greater wind gusts from a storm that has formed and has been named by the World Meteorological Organization. When a storm is farther from the coast, more locations at lower probabilities may be in play, but the picture will generally sharpen as the storm moves closer.

Fig. 3 | Evolution of highest wind gust threshold with meaningful probability in a replay of Hurricane Milton in 2024. Figure produced by Patrick Brown with Reask data and plotted with Matplotlib.

Resolution

As the storm’s life comes to an end, the National Hurricane Center will issue its final advisory and the settlement value will be published on the ForecastEx website here.

The settlement value is the single maximum calculated 3-second wind gust experienced in each location’s footprint over the entire course of the storm. If the value meets or exceeds a contract’s defined wind gust threshold at a location, the contract resolves YES on that threshold. 

Reask, a tropical cyclone risk analytics firm whose data underpins parametric insurance triggers across the global market, will serve as the resolution source via their Metryc product. The decision to resolve based on Reask calculations is international and driven by several factors. 

First, to back up for a moment, the simple, intuitive way to settle a wind contract is to use an anemometer. For example, a contract could settle based on airport weather station anemometers. The problem with that is that physical instruments have a well-documented tendency to fail precisely during the events these contracts are meant to measure, because power is lost, a communication link is damaged, or the instrument itself is destroyed. Airport observing stations in particular regularly go dark during major hurricanes, and post-storm analyses are routinely forced to note that the highest winds likely occurred after the instruments stopped reporting. A settlement source that systematically disappears during the peak of the storm is obviously not fit for purpose for contracts like these. 

Alternatives from sources like the National Hurricane Center’s Tropical Cyclone Reports carry authority but are unsuited for settling contracts like these because they are published many months after the storm and they are not standardized. They are narrative documents that compile whatever observations survived, at whatever points instruments happened to be present, in formats that vary from storm to storm.

Reask solves both reliability and timeliness issues with what they describe as virtual anemometers. Reask essentially calculates peak wind gusts for every 1 km × 1 km grid cell the storm affects, using the storm’s officially observed track and intensity, along with the known physics of how tropical cyclone wind fields interact with the specific topography and surface roughness of the land. 

Crucially, the virtual anemometer cannot lose power or be blown away. Using Reask allows these contracts to reflect our best estimate, based on the available observations, remote sensing, and known physics, of what actually happened at the location. This is the same reason the parametric insurance market adopts calculations like those produced in Metryc for payout triggers. 

Reask also calculates pre-event probabilities with a forecasting system called LiveCyc. LiveCyc generates on the order of a thousand plausible wind-field realizations for an approaching storm, conditioned on the National Hurricane Center forecast, and, from them, computes the probability that each location exceeds each threshold. 

ForecastEx will publish these location-level LiveCyc exceedance probability forecasts on its website as they are produced from Reask (here), so every participant has access to the same institutional-grade probabilistic baseline that, until now, was available only to institutions paying for it. 

Reask’s core methodology is well documented in peer-reviewed publications here, here, here, and here (see bibliography below). 

Fig. 4 | Settlement values for Hurricane Milton (2024) at reference locations in Florida. The virtual-anemometer approach yields a complete, consistent number for every location, and every listed contract resolves YES/NO against these values within days of the final advisory. In the case of the ‘where will the storm hit hardest’ version of the contract, St. Petersburg would have resolved YES and all other locations would have resolved NO. Figure produced by Patrick Brown with Reask data and plotted with Matplotlib.

Why 3-second gusts?

These contracts settle on the local 3-second peak gust, not the 1-minute maximum sustained wind used in the traditional Saffir-Simpson categories. 

This is a deliberate choice based on the evidence that structural damage scales much more closely with 3-second gusts than with 1-minute sustained winds. Engineering design codes (e.g., ASCE 7-95) are organized around gusts because roofs are lost, windows break, and trees are blown down during the strongest few seconds of wind. Furthermore, the wind a building experiences depends strongly on local terrain and surface roughness, which is what a 1 km terrain-interacting calculation captures.

For a rough scale, the World Meteorological Organization’s tropical-cyclone conversion guidance (Harper, Kepert & Ginger 2010, WMO TD-1555) suggests a 3-second gust is roughly 25% to 40% higher than the 1-minute sustained wind over land. 

It should also be noted that the 1-minute maximum sustained wind that is the basis for the storm’s assigned Saffir-Simpson category is the storm-wide maximum, which typically represents just a small patch over the ocean in one section of the eyewall and will often be highly unrepresentative of the winds experienced at any given location during landfall. 

Why these markets may produce the best hurricane wind forecast available

In addition to their risk hedging use cases (see series of writings listed at the bottom of this article), these markets have the potential to be highly valuable from an information discovery perspective. 

Based on accumulating evidence in both ForecastEx’s inaugural hurricane markets in 2025 and in ForecastEx’s temperature markets, the probabilities visible from these markets may constitute the best available location-level hurricane wind forecasts available, better than any single source, including the excellent data underlying them.

The mechanism by which this is possible is that prediction markets can aggregate conventional forecast information as inputs and convert them into refined probabilities through direct financial rewards for accuracy and direct financial penalties for inaccuracy. This creates a dual effect of attracting accurate individuals and systems into the market while deterring those who are inaccurate. People or systems that consistently make poor forecasts are strongly motivated to either improve or leave the market. 

Another advantage of prediction markets over traditional methods is their perpetually ‘live’ update frequency (as long as markets remain liquid). 

The underlying hurricane forecasts from both the National Hurricane Center and Reask’s own LiveCyc forecasts are constrained to update only about every six hours and also embed latency of another six hours or so coming from the cadence and workflow of incorporating the forecasts from the canonical global numerical weather prediction models. 

The status and expectations of a hurricane can change meaningfully within a six-hour window, so prediction markets are naturally more responsive to real-time observations and can produce extremely valuable changes in information in those gaps. 

These temporal gaps are also where speculative traders will likely seek an edge. Along those lines, Reask’s LiveCyc forecasts (posted live on ForecastEx’s website here) are centered on the National Hurricane Center’s central track forecast (with physically simulated along / across-track and intensity spread). 

The National Hurricane Center track forecast is very good, but it is one forecast, on a fixed schedule, and the National Hurricane Center, in its own 2025 retrospective, identified the Google DeepMind ensemble mean as having performed better on short-range track forecasts.

The bottom line for an interested trader is that the Reask LiveCyc probabilities represent extremely useful information for these contracts, but they are still anchored to a consensus baseline, updated only every six hours and thus leave substantial room for traders to find edges. 

The most open “Live cat” pricing to date

One motivation for designing contracts around ‘live’ hurricanes is that it mimics the ‘Live cat’ trading that reinsurers and specialist funds already engage in when a major storm is threatening. 

In principle, the structure of these contracts allows for an alternative means for a (re)insurer to hedge or transfer acute live hurricane exposure, with a firm that is over-exposed at a given location buying Yes on the higher gust thresholds and institutions seeking risk premium buying No. 

However, in the first iterations at least, exchange volumes are unlikely to be large enough for a reinsurer to lay off a meaningful share of risk using these contracts, so the most immediate value-add is not necessarily (re)insurance-scale risk hedging but rather bringing better price discovery to a previously opaque market.

In the existing live-cat market, Industry Loss Warranties (ILW) and short-dated retrocession are negotiated bilaterally through a small set of specialist brokers and trade over-the-counter (OTC). 

These Live Hurricane contracts reproduce the timing and the real-time character of this existing ‘Live cat’ trading, but rather than happening in the shadows, the activity happens on an open, regulated exchange, where the price is continuously visible, and anyone who qualifies can take either side. 

Thus, a broker, a reinsurer, or a smaller participant negotiating an over-the-counter trade as a storm approaches will now have open prices on wind-speed exceedances to anchor to. For a buyer of protection in particular, a visible, continuously updated market on an exchange serves to discipline the pricing of a much larger opaque market and can act as a check against being quoted too high. 

Further Reading on the Risk Hedging and Risk Transfer Applications of Hurricane Forecast Contracts

• Hurricane Forecast Contracts — how the count and landfall contracts work.
• Disaster Insurance Applications of Forecast Contracts — using landfall contracts as parametric cover.
• Hurricane Forecast Contracts as a Diversifying Asset in an Investment Portfolio — correlation properties against conventional assets.
• Bundling “NO” Hurricane Landfall Forecast Contracts to Reduce Downside Risk — constructing a bundle across locations.

Peer-Reviewed Journal References for Reask Methodology

• Loridan, T., Crompton, R. P., & Dubossarsky, E. (2017). A machine learning approach to modeling tropical cyclone wind field uncertainty. Monthly Weather Review, 145(8), 3203–3221. https://doi.org/10.1175/MWR-D-16-0429.1
• Loridan, T., & Bruneau, N. (2025). Reask UTC: A machine learning modeling framework to generate climate-connected tropical cyclone event sets globally. Natural Hazards and Earth System Sciences, 25(8), 2863–2884. https://doi.org/10.5194/nhess-25-2863-2025
• Mani, B., Loridan, T., Bruneau, N., Hannah, N., & Schmid, D. (2026). Reask Metryc: A probabilistic wind gust model for tropical cyclone event response with global coverage. Journal of Catastrophe Risk and Resilience, 4, Article 05. https://doi.org/10.63024/wnbk-07j1
• Loridan, T., Bruneau, N., Mani, B., Sampson, C., & Knaff, J. (2026). Local-scale wind forecasting for tropical cyclone early warnings. Bulletin of the American Meteorological Society. Advance online publication. https://doi.org/10.1175/BAMS-D-25-0182.1