The Cost of Borrowed Assumptions
Westren Capital · 18 June 2026 · 7 min read
Prediction markets have largely lived on the fringe of finance.
They were viewed as oddities – interesting during election times, perhaps even useful during major geopolitical events, but also fascinating in an academic setting as experimental devices to study collective intelligence. Observations would be made, predictions compared to polls and expert opinions, and the focus shifted somewhere else. The constant preoccupation centered around the question of whether prediction markets functioned.
This question is losing its appeal.
As liquidity increases, as participation widens, and as attention from institutions grows, a different issue is becoming increasingly relevant. Not the question of whether prediction markets work, but whether we know what we see when we examine prediction markets. For decades now, the academic study of finance has produced a powerful set of tools for interpreting markets. Concepts like liquidity, price impact, adverse selection, informed trading and market efficiency have all been studied in great depth in contexts where asset prices were linked to underlying future cash flows.
However, it seems to be clear from the outset that a prediction market belongs to this category as well, since it has an order book, market makers, liquidity providers, spreads and trading volume. This instrument displays numerous features associated with financial markets, but there is an important difference hidden behind the seemingly similar structure of the instruments.
This market does not attempt to estimate the price or the value of some asset. It attempts to predict the likelihood of an event. This might not seem evident at first sight, but the behavior of participants shows it clearly.
The contract on a prediction market always has a predetermined final state.
Any trade position will eventually reach certainty. If the contract trades at 0.60 today, then at some point in time it will be either 0 or 1.
This is an important property that alters the very essence of liquidity provisioning itself. In conventional financial markets, there are constant efforts to manage inventory on items with uncertain fundamental value. In prediction markets, it is managing inventory on probabilities that eventually resolve themselves into outcomes that are well known. It is a matter of different risk characteristics, different information dynamics and different participant behavior.
Nevertheless, much of the theoretical basis for analyzing these markets continues to derive from conventional finance.
That could become an increasingly expensive mistake to make.
The Risk of Analogy
One of the most intriguing findings from our analysis of prediction markets concerns market structure rather than forecast accuracy.
After analysing over 30 billion order-book events and more than 255 million on-chain transactions from Polymarket, we arrived at a counterintuitive conclusion. Standard tools applied extensively across market microstructure research managed to predict trade direction just barely better than by flipping a coin. Metrics considered indispensable by researchers could even have their signs flipped based on the source of order flow information public market data versus private blockchain settlement data.
Importantly, the implication of this result is far broader than the specific venue where it was uncovered.
Much like modern economics in general, market analysis rests on a series of assumptions that become so ingrained in the way we see the world that very few question their validity. Aggressive buyers and sellers can be differentiated. Price impact can be calculated. Effective spreads can be estimated. And informed investors can be identified.
However, any measurement system is necessarily based on some assumptions about the process of information getting into the market. When such assumptions prove inaccurate, the measurements themselves become unreliable. The problem faced by prediction markets is that most of the assumptions adopted from other financial systems were tailored for use in a radically different environment. Equity markets have grown up around instruments of unlimited duration and standardized accounting processes, while prediction markets are built around binary options with clearly defined expiration dates and hybrid structure combining off-chain order matching and on-chain settlement.
The outcome of all this is that prediction markets look similar to traditional financial markets at first glance, but are actually quite different when you take a closer look.
New types of assets tend to be evaluated using old frameworks for the time being, until their shortcomings come to light. While the questions remain pertinent, the techniques used for answering them need to be adapted.
Prediction markets may well be entering that phase.
How Information Becomes Price
The real significance lies in information.
Modern financial markets operate under the premise that prices are information aggregators. It is such an obvious concept that it is not questioned. Information is collected from the marketplace, incorporated into price, and translated into signals that no individual could otherwise create.
The function of prediction markets is essentially the same, albeit in a far more transparent manner.
It is not forecasts, discount factors or regime changes that are being predicted. Rather, individuals are stating their opinions about certain events. Elections, regulation, geopolitics and economic indicators become tradable predictions.
As such, prediction markets provide unique insight into the transmission of information through communities. The normal approach to interpreting markets tends to concentrate almost exclusively on the results. The trading price is 72%; thus, the probability is 72%.
However, the method behind the result is just as important.
Where is liquidity coming from? To what extent is there concentration in participation? How does the flow of information happen within the order book? What happens to liquidity as uncertainty starts being resolved?
These are issues which have received relatively little attention so far even though they are crucial for the quality of the signal.
According to our findings, liquidity in the prediction markets can be distributed in a way that many researchers have taken for granted. It does not seem to be concentrated in just one level, namely at the best bid/offer price but instead it seems to be layered along different price levels in the order book. It could appear that this is just an insignificant issue, but it may not be. Participants do not seem to trade in the same way that market makers hedge themselves against financial risk.
Increasingly, prediction markets are considered an alternative to polls, complementary to other models of forecasting, and an indicator of collective belief. These discussions have value. They also tend to center on the end point rather than on the means through which that endpoint is reached.
Markets don’t create information in isolation.
Markets create information through structure.
And structure creates signal.
End Notes
History is replete with instances of analytical tools lasting far longer than the context that spawned their creation. A helpful model morphs over time into received wisdom. The problem is not necessarily that the assumptions underlying the model are demonstrably incorrect, but that they retain enough verisimilitude to pass scrutiny.
Prediction markets may well be the first truly novel market structure that has developed to a significant degree since the digital era came into existence. Their rise will necessitate the reexamination of long-held assumptions that have remained largely untested for generations.
Perhaps the most notable change will not be their increasing size, but their ability to make us question assumptions that have been unchallenged too long.
Prediction markets certainly will have their futures influenced by issues of accuracy, increased liquidity, and expanded participation. But the really interesting question may reside below the surface.
Not what prediction markets forecast.
But rather how accurately we’ve been forecasting them.