Next Gen Prediction Markets w/ Ivo Sarjanovic

Ivo Sarjanovic co-founded Pascal after stints at Bridgewater, D. E. Shaw, and dYdX, where he served as CEO. His co-founder Matt has an HFT systems engineering background and was trading on first-generation crypto venues at age 19. Before building Pascal, the three-person founding team spent roughly three months running live trading systems on Kalshi and Polymarket to understand market structure from the inside. Their P&L was marginally positive but not exceptional, and their initial strategy of maximizing liquidity reward extraction across many markets was broadly a failure with a pennies-in-front-of-the-steamroller payout profile.

Pascal is a prediction market venue built on Solana using an off-chain matching engine with funds held in on-chain smart contracts, making it non-custodial. The team kept matching off-chain deliberately to preserve traditional HFT sequencer guarantees and avoid MEV semantics. Pascal charges fees three to five times lower than Polymarket depending on market category, pays half of the taker fee to the maker atomically on every fill, and its net exchange fee is between one sixth and one tenth of Polymarket's fee.

The team's time trading on existing venues produced specific structural critiques. Polymarket's matching engine takes two signed messages with nonces from each side of a trade and sends fill notifications over a websocket before anything is finalized on-chain. Because Polygon takes approximately eight seconds to finalize, a trader can submit a different transaction with the same nonce and higher gas directly to the Polygon sequencer, invalidating a matched trade before it lands on-chain. This nonce attack causes fills to disappear at the worst possible moment for the liquidity provider and constitutes adverse selection. Polymarket did not send an affirmative cancellation message when a trade was busted, and trades had three levels of finality where traders would sometimes receive the first two confirmations but never the final on-chain confirmation. Polymarket has since hired good people and upgraded its smart contracts to largely fix the exploit. On Kalshi, queue priority is the most important edge factor because tick sizes are large and order flow is soft, making front-of-queue positions highly profitable against retail flow. Polymarket's liquidity rewards system is modeled on the dYdX formula, and in practice market makers want to be at the back of the queue so they are never filled but still collect rewards, adding zero value to the venue.

Polymarket's yes-no token architecture requires a trader holding 100 YES tokens who wants to be 100 NO to buy 200 NO tokens and then merge them, making liquidity provision highly capital intensive. Pascal eliminates this structure so that selling 200 contracts when long 100 results in being 100 short with no merge step. Pascal also supports scalar markets that can settle to any value between zero and one rather than only binary outcomes, enabling linear exposure to events such as IPO price ranges or rainfall levels.

The team concluded that successful small prediction market trading teams are modelers first and engineers second. One of their angel investors is a two-man team whose founder is a computational chemist who models weather and sets price levels manually without systematic engineering. Pascal aqua-hired one of the successful small trading teams active on Polymarket and Kalshi after concluding that lacking price discovery expertise makes prediction market making a very difficult game. Serious firms like SIG will not market make on small viral prediction markets, and large price swings or one-sided money flows in those low-volume markets almost certainly reflect inside information.

Pascal operates an OTC desk aimed at counterparties with genuine economic risks to hedge. Despite this ambition, real risk-exchange volume remains thin, and in the team's estimation Kalshi has printed fewer than ten genuine risk-exchange block trades. The speakers believe real economy hedging will first emerge adjacent to popular retail categories like sports and weather. A concrete example involves large data centers and neoclouds that are increasingly preferring derivatives contracts over traditional flood insurance, with payouts triggered by measured water depth at thresholds of one inch, six inches, and twelve inches rather than by a multi-month claims assessment process covering up to one hundred million dollars of capital assets. Neoclouds operate with extreme leverage, borrowing to buy GPUs and collateralizing loans against them, meaning a large insurance payout is not a useful hedge because the business cannot survive the downtime that triggers it, making fast-settling event contracts structurally more attractive.

The speakers are skeptical of the exchange model for purely retail gambling audiences. In the UK, sportsbooks outcompeted betting exchanges because the full lifetime value of a losing user is captured by the sportsbook rather than split with market makers, and sports bettors empirically do not care about price or frictions. Kalshi's current success is characterized largely as regulatory arbitrage via CFTC oversight, and the speakers predict that if the regulatory environment does not improve, sportsbooks will eventually outcompete it. Pascal is also building a leverage layer separate from its core matching engine to address the capital efficiency problem inherent in fully collateralized event contracts, drawing an analogy to how early perp exchanges like BitMEX had to build leverage provision into the exchange itself because no external prime brokerage layer existed. The speakers acknowledge this is not an easy problem and many teams are attempting it in different ways.