lattice
Prediction markets are powerful, but they were built for isolated outcomes. Once users start combining related events, traditional pricing breaks because it assumes every leg is independent. Lattice fixes that. Lattice is a correlation-aware structured prediction market that uses semantic analysis and Gaussian-copula mathematics to price multi-market stacks more accurately. Users can build, share, and copy structured theses, while liquidity providers earn through a protected USDC vault. Every quote is signed, risk-checked, and executed on-chain through a gas-sponsored smart account.
Video
Công nghệ sử dụng
Sự miêu tả
Lattice is a correlation-aware structured prediction market that replaces naive parlay multiplication with defensible probabilistic pricing.
Traditional multi-outcome products estimate a stack using:
P(stack) = P₁ × P₂ × ... × Pₙ
That formula only works when every outcome is independent. In prediction markets, that is rarely the case. Elections, interest rates, oil, wars, sports tournaments, and crypto prices often share the same entities, timelines, catalysts, or logical relationships. Treating these outcomes as independent can create inflated payouts and allow sophisticated traders to extract value from liquidity providers.
Lattice addresses this with a deterministic five-layer semantic correlation engine. Every pair of legs is evaluated for lexical overlap, shared entities, common themes, event and time proximity, and logical alignment or contradiction. These signals are converted into signed pairwise correlation coefficients, ρᵢⱼ, and assembled into a valid correlation matrix.
The engine then uses Gaussian-copula joint pricing. Marginal market probabilities are transformed into normal thresholds, correlated through Cholesky decomposition, and evaluated with antithetic Halton quasi-Monte Carlo sampling. The resulting joint probability is constrained by mathematically valid Fréchet bounds, preventing impossible probability outputs.
Lattice therefore prices each stack using:
Effective joint probability = Cρ(P₁, P₂, ... Pₙ)
and calculates the payout as:
Payout multiple = (1 − protocol risk premium) / effective joint probability
The gap between the independence baseline and the effective probability becomes a visible correlation adjustment. Positively correlated legs receive compressed payouts, genuinely diversified stacks stay closer to independent pricing, and mutually exclusive outcomes are recognized rather than incorrectly treated as ordinary positive correlation. Extreme long-tail payouts are capped at 100x to protect vault solvency.
This is not AI generating arbitrary odds. The pricing engine is deterministic, testable, bounded, and reproducible. AI is used only as an explanation layer, showing users why legs are related, which correlation signals affected the quote, and how the final payout differs from naive multiplication.
Every executable quote is bound to its recipient, stake, return, expiry, ordered legs, metadata, and Robinhood Chain domain through EIP-712. On-chain contracts independently verify the signature, leg hash, quote expiry, market status, stake limits, payout limits, liquidity availability, and replay protection before any capital moves.
Users execute stacks through gas-sponsored smart accounts and receive on-chain position receipts, reusable bet codes, shareable cards, and one-click copying. Liquidity providers deposit USDC into a vault that reserves payout exposure and enforces delayed withdrawals for bank-run protection.
Lattice combines institutional-style correlation modelling, transparent risk controls, DeFi liquidity, and social prediction-market distribution in one product.