Charlotte Nimmo on Why Prediction Markets Disagree on Price

Prediction markets let people trade on whether a real-world event will happen. Each contract asks a yes-or-no question and settles at one dollar for yes or nothing for no. According to CertiK's 2026 Skynet Prediction Markets Report, global trading volume rose fourfold from 15.8 billion dollars in 2024 to 63.5 billion dollars in 2025, and the same outcome can be priced differently from one venue to the next. iPredicta launched in July 2026 as a comparison platform for US prediction markets, putting live prices from several venues, including Polymarket, side by side and earning affiliate fees when users go on to trade. Its founder and chief executive, Charlotte Nimmo, previously spent four years running Best of Bets, an affiliate betting platform, after more than a decade leading her own PR agency across London, Stockholm and Geneva. She answered The Fintech Times' questions in writing on what carries over from betting, where the line with gambling sits, and why the company's AI is allowed to write the words but never the numbers. Prediction markets found a following during the 2024 US presidential election, Nimmo said, but 2025 was the year the category moved beyond politics. 'Sports, crypto, economics and other real-world markets broadened the audience, and annual volume rose to $63.5 billion, according to CertiK.' She points to Bernstein's projection of around 240 billion dollars in volume for 2026 and one trillion dollars by 2030, and to a Pew Research Center analysis of data from The Block which found that combined monthly volume on Kalshi and Polymarket rose from less than five billion dollars in September 2025 to about 24 billion dollars in April 2026. For her, the clearest sign that the category is going mainstream is who is entering it. 'FIFA appointed ADI Predictstreet as its first official prediction market partner, while Robinhood, DraftKings and Underdog have all moved into prediction markets. Institutions are also starting to use the data itself as a live measure of sentiment. I think the growth is durable, but regulation now has to keep pace.' One question, two prices The premise of a comparison site is that the same outcome trades at different prices on different venues. Nimmo's explanation is that it is usually not quite the same market. A contract on the same event can trade at 52 cents on one platform and 56 cents on another, and that is not a glitch. 'It simply comes down to different liquidity, different users, different fee structures and different market structures.' A six-cent gap between Kalshi and Polymarket on the same question can look like an arbitrage opportunity. 'It may be, but only if the contracts resolve on genuinely equivalent terms and the price difference remains after fees.' The two platforms are run differently and take different considerations into account, and that alone can produce a gap. 'The bottom line is that Kalshi and Polymarket are not the same market. They are separate exchanges with separate order books, separate user bases and separate market dynamics. The fact that they are asking the same question does not mean they will arrive at the same price, any more than two different stock exchanges would price a cross-listed equity identically at every moment.' Every price on every platform, she said, reflects the balance of buying and selling pressure on that exchange at that moment, and understanding the mechanics behind the discrepancy tells you what you are actually comparing. From punters to positions Much of the affiliate model carries over from betting: offers and promotions for new users in return for a one-off fee or a revenue share, and insight, education and data for the markets covered. The difference is breadth. 'One of the main differences is that, with prediction markets, we cover much more than just sports. Users can trade on all sorts of real-world events, including elections, economic data, tech launches, crypto prices, cultural events and much more.' That means producing content for each of those markets, but it also widens the audience, 'including, from what I have seen, a lot more women'. The user is different too. 'Many think of themselves as traders rather than punters. They want depth, spread, settlement wording and time to expiry. They can also exit before the event resolves, which turns it from a bet placed into a position held. That is a different product and a different duty of care.' Hedge or entertainment Asked where the line falls between prediction markets as information tools and as gambling products, Nimmo said the line is not drawn by the instrument, which is exactly what makes it difficult. A contract on the next inflation print can be a hedge for a business exposed to inflation through its costs or revenues, and entertainment for someone at home. 'Identical contract, two different purposes, and no drafting separates them at the point of sale.' One test is economic purpose: whether the contract lets somebody transfer real-world risk, and whether there is a plausible population who would use it that way. 'In my view, contracts on rates, elections and geopolitics broadly pass that test. Most sports contracts do not, and I say that as someone whose business benefits from sports volume.' Her criticism of the industry is that it has spent its energy arguing about the label. 'Whether you call it a derivative or a bet matters far less than whether a retail user gets the same protection either way. Age limits, affordability checks, advertising standards, and self-exclusion that works across both regimes rather than stopping at the boundary.' Her position for regulators is, in her words, straightforward to state and awkward to implement. 'Settle the jurisdictional question however it needs settling, but do not let the user-facing protections depend on the answer. Nobody should end up with weaker safeguards because of which regulator won.' Words, not numbers iPredicta's AI layer explains each market in plain English but, in the company's own phrase, produces the words and not the numbers. Nimmo's reason for drawing the line there is blunt. 'Because a language model is a good writer and a bad source. It will produce a fluent, confident, correctly formatted number that it had no way of knowing. In a financial context, that is not a typo. It is something a reader might act on.' Every figure that reaches a reader comes from the venue's own data at the moment of writing and is checked against live data again before anything is published. The model is left with the part it is good at: what the market is, how it resolves, what has moved, and what would have to be true for the current price to be right. The risks, she said, are not the ones people expect, 'because the failure mode is usually plausible wrongness'. Copy can read well while describing a market that does not exist, a headline figure can quietly contradict the table beneath it, a price can attach to the wrong outcome on a multi-outcome market, and resolution criteria can be paraphrased so smoothly that they no longer match the contract. 'We handle that with automated checks that block on figure mismatches and unsourced claims. A person signs off every piece, so nothing publishes on its own. As AI makes authoritative-looking commentary almost free to produce, provenance becomes the thing worth paying for.' Fragment first, consolidate later Nimmo expects prediction markets to fragment before they consolidate, with more venues, wider distribution and regional products built around local licensing rather than one global venue serving everybody. 'That makes a line such as 'the market says 34 per cent' much less useful. You need to know which market is being quoted and what that contract actually settles on.' The second shift is the marginal user. 'Not a crypto native and not a quant. Someone who saw a probability quoted on the news and wants to understand what it means. That is a comprehension problem before it is a trading problem, and most of the industry is not built for it.' The third is a category mix moving toward macro and economics, which she describes as where institutional capital goes and where the product looks least like betting. iPredicta's own position in that landscape is deliberately narrow. 'We are not a venue nor do we run markets, hold client money or take a position. That is deliberate. We are the layer that shows you where an outcome is priced across venues, what each contract actually settles on, and whether a gap reflects a genuine price difference or simply two different questions. Comparison businesses do well when a market fragments. Ours is fragmenting.' At launch, iPredicta tracks dozens of live markets across two venues, with prices updating automatically as markets move, and scores each market on liquidity and uncertainty, with alerts when prices change.