Which of Polymarket's multi-market pathologies come from discretising a continuous quantity versus the binary architecture itself? By splitting 18,863 events into continuous (price, margin, temperature buckets) and categorical (candidates, teams) slices and re-running the v1 analysis, functionSPACE shows concentration is architecture-wide, ghost markets are largely a categorical phenomenon, and a continuous-distribution primitive is a sharper fix than v1 suggested.

Let's find out how Polymarket handles complex questions by breaking them into multiple yes/no contracts. By examining metadata from the Gamma API, functionSPACE argues that this "fragmented" approach creates a "resolution gap" where liquidity fails to spread evenly across all outcomes.

Polymarket traders have an inherent psychological bias toward "Yes" outcomes. By analyzing over 7,000 events, the researchers discovered that the platform’s editorial tendency to frame questions around dramatic, unlikely scenarios (e.g., "Will a specific event happen?") naturally makes the "Yes" token a cheap long-shot. Their data reveals that traders don't actually care about the "Yes" label; they simply gravity toward cheaper tokens regardless of their name. Consequently, what appears to be a behavioral bias is actually a structural illusion created by price sensitivity and the way markets are designed, where the "No" outcome is the default reality for most unlikely events.

We argue that prediction market TAM should include the supply side: as the cost of producing real-time probability estimates collapses, the addressable market extends beyond trading volume to every decision that benefits from better forecasts.

The smart-money-vs-dumb-money framing misreads how prediction markets actually work. Drawing on three pieces of forecasting research (NBER, Wharton/INSEAD BIN model, Kapoor and Wilde on cognitive search), functionSPACE argues noise is a structural requirement, not a bug - noisy traders fund the probability space that informed participants sharpen as evidence arrives - and that continuous probability markets harness noise as the shape of the curve rather than treating it as a cost.

functionSPACE argues the core contradiction in permissionless prediction markets - decentralised trade, centralised settlement - only resolves when the oracle itself becomes a market, where disagreement becomes price discovery rather than a governance game.

Prediction markets need to agree on what actually happened - a state confirmation problem. functionSPACE compares delegated approaches (Kalshi) against collective ones (UMA, Reality.eth, Chainlink, AI oracles) and argues that a market-led confirmation surface, where capital replaces voting weight and outcomes sit on numerical ranges, is the only mechanism auditable at the incentive layer.

Prediction markets look marginal at less than 0.5% of DeFi volume, but reframed through forecasting, insurance, and hedging they look early rather than small. functionSPACE maps the emerging four-layer prediction economy - infrastructure, venues, builders, and the interaction layer - and argues the next phase of growth won't come from new venues but from where people encounter the ability to express belief.

Beliefs aren't binary - they're directional, bracketed, distributional. functionSPACE argues that binary event contracts force 1-bit belief expression and fragment capital by factors of up to 256, while a shared continuous probability surface lets traders express uncertainty directly, reward calibration, and turn beliefs into composable information assets.