How Predictive Are Prediction Markets?

How Predictive Are Prediction Markets?

Luana Lopes Lara, Kalshi’s co-founder discusses what she’s learned about the exchange’s forecasting power since it launched last summer

When Tarek Mansour and I started Kalshi, we had a vision of an “everything market,” an exchange where people could hedge, forecast and trade on the things they were talking about. Forecasting was always an important part of the vision: Both of us had read the classics from Philip Tetlock and Robin Hanson, and one of us had done research in forecasting using statistics and machine learning.

Originally published In Luckbox Magazine. Subscribe for free at

We dove deeper into the literature on forecasting, reading papers from economists, such as Justin Wolfers and Eric Zitzewitz, that showed why prediction markets were theoretically optimal information aggregators, and psychology papers that showed how the markets could fail.

In the process of building the exchange, we used every shred of this acquired knowledge to choose the order book model, design the fee structure and craft the trading mechanics—all with an eye toward maximizing the forecasting power of our markets.

Let’s take an example market category, COVID-19, and see how it has performed in the wild.

One of the most interesting COVID-19 markets is one that hasn’t yet settled: VOHC-001. This market asks if the CDC will identify a “variant of high consequence” (VOHC) by March 1, 2022. Vaccines and therapeutics are ineffective against VOHCs, and no variant has yet been labeled one. The Delta variant, for instance, is a “variant of concern,” which is one tier below.

From the market’s launch on Sept. 6 to the time of this writing, traders have indicated that the probability of a VOHC emerging is at least 20%. Already, this is notable because predictions from news media, politicians and even scientists have been—and continue to be—far more optimistic about the continued usefulness of vaccines and COVID prevention measures.

A spike appeared in the “YES” probability on Nov. 25. What changed? That was the day South Africa reported news of the new, highly transmissible Omicron variant. There wasn’t much information about the variant at the time, except that the World Health Organization quickly labeled it a “variant of concern.” Traders drove yes-side prices up to 50%.

In the days after, media reports questioned the new virus’s transmissibility and lethality, causing the probability of “YES” to swing rapidly before eventually settling around 40%. By early December, as even more was learned about Omicron, the probability was back within the mid-20% range.

The market is still predicting a higher probability of a VOHC than is reported in the news or suggested by scientists. What does this mean? Prediction markets can produce very different and more frequently updated forecasts than other sources.

Kalshi has another series of markets, called VAXX, forecasting the number of vaccinated Americans. The underlying data source is published daily by the Centers for Disease Control, and it’s updated with a one-day lag.

VAXX markets were structured to fit a weekly schedule, which means that if the market is functioning properly, it will become more determined over time as the possible variation in the weekly number is reduced by the trickle of daily data points.

Over the weekly markets’ lifetimes, prediction error drops significantly as the market nears resolution, suggesting the markets responded extremely quickly to new information and updated probabilities without succumbing to narratives or biases prevalent in other sources.

COVID-19 is just one example, but the rest of Kalshi’s markets work the same way. They’re constantly updating, taking into account new information and providing facts-first forecasts.

Our experience with Kalshi has reaffirmed our belief in the power of prediction markets, and we’re excited to share this powerful tool with the world.

Luana Lopes Lara co-founded Kalshi, the first-ever federally regulated events-contract prediction market.@luanalopeslara2

Prediction Markets

Augur was developed by Forecast Foundation as a global no-limit decentralized prediction market platform built on the ethereum blockchain. Betfair, a London-based online gambling company, operates the world’s largest online betting exchange and hosts prediction markets on domestic and international politics.

Kalshi, a new federalla regulated prediction market based in the United States, began its beta in July 2021 with markets based on COVID-19, climate, politics and pop culture.

Nadex (North American Derivatives Exchange) is a regulated financial exchange, specializes in short-term binary options and spreads. It recently introduced a suite of prediction markets for forecasting economic events.

Polymarket, a prediction market platform built on the Polygon blockchain, uses USD Coin (USDC), a cryptocurrency stablecoin, to make trades.

PredictIt, a New Zealand-based prediction market, was launched as a research project in 2014 with a no-action letter from the Commodity Futures Trading Commission. It emphasizes politically themed markets.

Originally published In Luckbox Magazine. Subscribe for free at

Source link