Vegas in the Rear View Mirror

March 16, 2026

Prediction markets need to leave Vegas behind. In this article, we’ll explain what we’re building at Talarion, and why. But first, a little context.

Trading on Prediction Markets is a Losing Game

The average prediction market trader loses money. Why? Because exchanges charge fees, meaning derivatives markets aren’t zero-sum — they’re negative-sum. Every contract that trades at price $p$ on Kalshi generates $f = 0.0875 \cdot p \cdot (1 - p)$ in fees. In the worst case (a contract trading at $50$ ¢) that’s nearly $2.2\%$ . In the average case, it’s close to $1.6\%$ . It’s worth highlighting that $80\%$ of that fee is paid by the taker. Furthermore, consider the incentives for an institutional market maker on Kalshi, such as SIG or Jump Trading. They’re there to make money. Lots of it. Having the taker pay the lion’s share of the fees isn’t enough: for market makers to stick around, their trades need to make money on average. And that means that you, dear reader, need to lose more than $1.6\%$ . You are the fish, and your willingness to lose money is quite literally the product.

Why Does the Stock Market Go Up?

It’s a simple question, but it cuts straight to the core difference between capital markets (stocks and bonds) and derivatives markets (options, futures, prediction markets). When you buy a stock, you might be buying it from a short seller who thinks it will go down in price, but you might not be. You might be buying it from a company that is raising money in order to do something. If it does something good that people like, you will earn a return on your investment. The money you make is what we might affectionately refer to as value creation: you made it, but that doesn’t mean that someone else lost it. Notably, this is not a feature of derivatives markets. Every dollar that SIG makes on Kalshi is a dollar that retail traders have lost.

No Free Lunch

Any derivatives exchange which promises profits to its users is selling them snake oil. Futures markets were born out of the need to hedge against changes in the price of agricultural products. If your counterparty has some intrinsic reason to need to hedge, then plausibly you’re making a positive expectancy trade. If not, well, somebody is eating somebody’s lunch. For the lunch to be free, the underlying market needs to be positive-sum, which means a real person doing real stuff is benefiting from the market’s existence.

Is There Any Value Creation?

In the case of capital markets, the real value creation occurs when hardworking employees at ordinary companies show up for work. Luckily for prediction market traders, they are also creating value, just of a rather different sort. Maybe the price is the product. The information that Kalshi and Polymarket aggregate in the form of market prices may well be worth more than the fee revenue that they currently generate, but we have no idea, because nobody pays for it. And why should they? It’s sitting there on an open orderbook. Traders contribute valuable information, but on average they pay for the privilege of doing so.

For the world to be different, we need to build an Information Marketplace.

Only Some Questions Are Worth Answering

As of the time of writing, more than $90\%$ of Kalshi trading volume is sports. These markets generate information which is intrinsically valuable to media platforms and arguably to sports traders, if and when they can find a counterparty that hasn’t bothered to seek it out themselves. Sports betting isn’t positive-sum and it probably never will be, for the simple reason that sports bettors aren’t typically contributing information that anyone ought to pay for. For an Information Marketplace to exist, we need to prioritize questions that are worth answering.

Sell What You Know, Earn What You Can

At Talarion, we let you trade on whatever you want. We have the mother of all adverse selection problems on our hands. But the point is to lose money on the trades in order to generate valuable information. We set a prior and a spread on the markets we show you, because we won’t pay you for what we already know. We’ll pay for what we don’t. The information that you have and we don’t — that’s the product.

What’s Next

We’ll be writing about:

Quantifying the value of information
Practical strategies for building on the Talarion API
The importance of LLM advancements for judgmental forecasting

Stay tuned.