Intro to Prediction Markets, Part 1: Comparison to Other Markets, Especially Derivatives

Prediction markets share many features of other financial derivatives. First among them is the idea of an underlying quantity, from which the contract's value is derived. Ultimately, profit and loss will be dictated by the trader's ability to forecast this underlying.

However, there are many differences between prediction markets and other derivatives that will be examined in this post. Many of them are systemic, those that depend on the very mechanisms of the markets in themselves, but some are idiosyncratic, results of present market maturity.

Systemic Differences

Fundamentally, the exact valuation from a prediction differs from other financial derivatives. Consider a market where we expect an underlying to have value $S$ at expiry, which is more than strike $K$. The value of a long futures contract at expiry is $S-K$, and for a long call option the value is $\max\{0,S-K\}$; for an analogous prediction market contract, the value would be $\Theta(S-K),$ where $\Theta$ is the Heaviside step function. You are only correct in that you predicted that $S>K$ and gained no profit for the magnitude of your correctness. As previously mentioned, this type of prediction market is a digital call option, but the discrete valuation remains a feature in all prediction markets, with several immediate consequences.

The first consequence is deadline convergence: All contracts will be priced at either 0 or 1 at the end; if there is a continuous underlying, the value function becomes discontinuous at the strike. For options or futures contracts, small fluctuations in $S$ result in small profit changes, not so for prediction markets. Secondly, deadline convergence dramatically affects mark-to-market accounting, because all profits and loss can flip so rapidly.

Because of the binary nature of returns, many classical assessment tools must be reconsidered:

• Skewness becomes quite important, by design. e.g. a trade at $p=0.05$ either loses 100% or wins 1900%, making it very right-skewed.
• The price domain is a bounded (0,1). This and the above point about return size will suggest a change towards using log-odds as a natural coordinate system.
• Volatility $\sigma^2(\tau) \propto p(1-p)/\tau;$ deadline convergence forces $\tau\to0$ but also $p(1-p)\to 0,$ but the ratio remains an interesting quantity.
• Continuous distribution metrics for returns like VaR or the Sharpe ratio break.
• No value beyond the raw probability: If your estimate of $p^*$ is better than $p$ that is your edge.

Future blog posts will go into greater detail about these and other quantitative features that must be understood before trading on prediction markets.

Furthermore, the YES/NO nature of contracts results in what amounts to a zero-sum game: my YES winning means your NO loses. Fees and vig make them a negative sum game, and the empirical distribution finds that most people lose. The correct reference frame for intra-market skill is therefore gross return. However, one must still beat the risk-free rate before deploying capital (note that in normal markets the risk-free rate is the correct reference frame for intra-market skill). This produces a ladder of different benchmarks for a prediction market strategy to clear. Beating equity markets (primary trading source) > Beating risk free rate (deploy capital) > Breaking even (strategy works) > beating the median (Still losing money)

Another big difference between financial derivatives and prediction markets is oracle risk. With a normal derivative, the profit from an executed contract comes from buy at price $K$ and then selling at $S$, and this value comes from the market itself. For prediction markets, each contract has a resolution mechanism within to evaluate itself, and understanding the specifics of the resolution are important. This becomes especially important when doing cross-venue strategies, as two markets that nominally have the same underlying may differ in how they measure it. For example, a city with two temperature markets may be reading from two different thermometers. Mechanisms exist to dispute the resolution of a contract, but that is an irregular situation.

The nature of the various oracles also presents another difference: resolution can be event-driven, not calendar based, making their analysis trickier. For example, consider an annual election market strategy wins 0.3846% for a week of locked up capital. Saying that it would annualize to 20% is correct insofar as a that would happen IF elections were every week, but they don't so it is a bad metric. The development costs would probably require much greater returns. By contrast, 15 minute crypto market strategies that run continuously can be analyzed by more traditional means.

Present Idiosyncrasies

The above differences are dependent upon the nature of prediction markets and other derivatives. However, there are also present differences that one should be aware of broadly before entering markets. Parts of these may not be true by the time you read this, so do look into it yourself.

Prediction markets are smaller than other markets by a large amount, and so counterparties are fewer and more identifiable. Most price discovery is driven by skilled winners (Nechepurenko, arXiv:2605.02287). with better information who serve counterparty to many retail traders, producing adverse selection in participation, not just in working the orderbook.

Liquidity is often a concern, with some markets working on very small amounts of money. So even if one has a clear edge on a small market, there may not be a counterparty available. This is enough of a problem that venues often have liquidity rewards programs to incentivize market makers to operate on the markets such as the Polymarket -1.4% maker rebate. Liquitdity will also be the topic of a future blog post.

Prediction markets also exist in a state of ever-changing legal frameworks. In the US, different agencies and legislative bodies at both the state and federal level are processing these legality of prediction markets differently. Regardless of present or future legality, the underlying mathematics will not change, but fee structures, resolution mechanisms, and trading mechanisms could. We will continue talking about these topics in future blog posts.