Prediction markets provide a venue where anyone can numerically prove whether they have differentiated insights that the market is not taking into account. Claim markets apply this to public figures: we translate what people say into prediction market positions, providing an objective accuracy measurement across domains. This helps the public identify who has signal (positive contributor) and who is noise (negative contributor).
The result is an equilibrium where people are incentivized to make more accurate predictions. Public figures also get an early signal of what they're uniquely good at. Someone who is outperforming the market on Chinese geopolitics but not on European geopolitics gets a market signal that they might have a better macro understanding of China than Europe.
People, not models, live at the frontier of the most interesting context-specific knowledge of a given domain. By tracking real public figures rather than abstract models, claim markets measure something people actually care about: who should you listen to, and on what topics?
Vistadex is a prediction markets exchange built on Solana. Prediction markets let you trade on the outcome of real-world events. Each market has two sides: YES (the event will happen) and NO (the event won't happen). Contract prices range from 1¢ to 99¢, and the price of YES + NO always equals $1.00. If you hold the winning side when the market settles, you receive $1.00 per contract. If you hold the losing side, you receive $0.00.
Vistadex is a full-stack platform that allows you to trade all events at the best price, currently integrated with all Polymarket and Kalshi markets. Users trade with real money on Solana. For more on how Vistadex works, read the docs here.
Unlike standard Vistadex trading, claim markets use paper money. Each tracked person starts with a virtual balance and all trades execute at the market's current midpoint price. Positions are marked to market continuously, and the leaderboard ranks tracked people by total account value (cash + positions). Tracked people that identify mispriced markets or correctly predict outcomes will generally have higher account values than those that don't.
Paper trading makes the benchmark execution-agnostic. We care about whether a tracked person's predictions are directionally correct, not whether they can get the best fill on an exchange. All paper trades execute at the market's current midpoint price using a consistent snapshot, so every tracked person is measured on the same terms.
Each tracked person starts with $10,000 in paper money with a fixed trade size of $100. Positions are marked to market continuously using current prices, so account values reflect what a tracked person's portfolio is actually worth at any given moment. This creates a verifiable record of each claim's implied accuracy.
Each tracked person has a dedicated listener on their X feed. We ingest all posts and extract any relevant claims. Beyond X, we also scrape their broader internet presence including blogs, podcasts, interviews, and public appearances to find anywhere they made a prediction with a related prediction market.
LLM-powered extraction identifies forward-looking predictions including explicit bets, confident assertions, probabilistic forecasts, conditional predictions, and directional claims. Vague or unfalsifiable predictions are filtered out, and duplicate claims are automatically removed using content-based deduplication.
Actionable claims are automatically matched to existing prediction markets. An LLM generates search queries from the claim's core topic, searches for candidate markets, and selects the best match based on how directly the market question relates to the claim's prediction. Claims without a suitable market match are flagged and skipped.
Trading is scoped to the categories and tags that each tracked person is best known for. Trust is implicitly understood to be across a specific vector: you wouldn't trust a tech reviewer for medical advice and you wouldn't trust your doctor for tech advice. By limiting the scope of tradable markets to what each tracked person is known for, we get better signal from noise. Certain market types are also blocked, such as short-term crypto price movements and in-game sports trading, where speed-based pipelines dominate and claims provide no edge.
The default leaderboard ranks tracked people by paper PnL: realized plus unrealized profit or loss from their automated claim-market positions. Account balance remains visible as cash plus the current mark-to-market value of open positions.
Signal ranks paper-trade outcome value over entry expectation. Finalized payouts count at full weight. Active positions use discounted mark-to-market value until markets settle, so current market movement can contribute without being treated like final truth.
Signal rank uses the point estimate. Confidence bounds, scored counts, active mark-to-market counts, and unscored settled counts are shown as context, not as the rank key.
Markets resolve based on publicly verifiable outcomes. Each market specifies its resolution criteria upfront, including the data source and the conditions under which the claim is considered true or false. Resolution is binary: the claim either came true or it didn't.
When a market settles, tracked people holding the winning side receive $1.00 per contract and the losing side receives $0.00. The realized PnL from settlement is reflected in their account value and track record. Unsettled positions remain marked to market at current prices until resolution.
Not all predictions can be captured. Nuanced or heavily conditional claims may not map cleanly to a binary prediction market. Claims without a matching market are skipped entirely, which means some predictions go untracked.
LLM-based extraction is imperfect. The pipeline may misinterpret sarcasm, hypotheticals, or restatements of someone else's view as a genuine prediction. We filter aggressively, but some false positives and false negatives are inevitable.
Paper trading removes execution risk but also removes the signal that comes from having real capital at stake. A tracked person's paper PnL reflects directional accuracy, not trading skill.
Active Signal partly depends on current market prices before settlement. Thin or noisy markets can move the score before the final outcome is known, which is why active mark-to-market contribution is discounted.
Selling is not yet live. If a person walks back their view, this is not reflected in a sold position. Instead, the updated stance is captured as a new position, meaning both the original and revised views may be tracked simultaneously.
Claim markets is an ongoing project. Here's what we're working toward:
For any questions, comments, or feedback, please reach out to founders@capitola.xyz.