# Prediction market methodology: signals, vetoes, caps, and explanations Published: 2026-06-17 Updated: 2026-06-17 Canonical: https://orrery.me/research/prediction-market-methodology Markdown: https://orrery.me/research/prediction-market-methodology/markdown A methodology is not a magic score. It is a contract for what the system may observe, what can block a paper intent, and what must be explained later. ## Direct answer - The methodology starts with source-backed observations, not model authority. - Vetoes matter as much as signals because many markets are better ignored. - Paper evidence is the proving ground; live execution remains separately gated. - Every useful methodology output should be explainable before and after the outcome. ## Direct answer A prediction market methodology should say what the system is allowed to observe, which risks can veto an action, how costs and liquidity affect sizing, how outcomes are labeled, and how the system explains itself later. Orrery's methodology is verification-first. It is source-backed, read-only at the public boundary, paper-only unless explicitly approved otherwise, and built to explain why a market entered review, why it was vetoed or approved for paper, and what was learned after resolution. ## Signals are observations Signals detect market conditions: movement, divergence, flow, microstructure quality, semantic risk, resolution risk, drift, and no-trade indicators. They do not become authority by themselves. A signal should carry kind, evidence, feature values, freshness, source, and reason codes. If those fields are missing, the system may still produce a score, but it cannot explain itself. ## Vetoes are part of the alpha Many prediction markets should not be touched even when they look interesting. Liquidity can be thin, spreads can erase edge, source wording can be ambiguous, wallet flow can be manipulative, or the market can be stale. That is why the methodology treats vetoes as first-class. A no-trade decision is not failure. It is the system saying the evidence is not clean enough. - Liquidity veto: the book cannot support the assumed entry. - Resolution veto: the contract wording or source path is unclear. - Manipulation veto: flow looks concentrated or suspicious. - Cost veto: spread, impact, or fees consume the apparent edge. - Drift veto: recent paper evidence suggests the method is degrading. ## Paper trading is the validation layer A methodology should not jump from readiness to authority. Paper trading records the same decision flow without live orders: opportunity, intent or veto, paper fill, costs, outcome label, explanation, and learnable/dirty label. This creates the evidence needed for trust. A system that cannot explain paper wins and losses should not be trusted with live execution. ## Shadow ML can challenge, not command Machine learning can be useful when it proposes alternative probabilities, target actions, or abstention choices. But it should remain shadow-only until it beats the source-backed methodology under off-policy evaluation, coverage, calibration, drawdown, and drift checks. The safest architecture is challenger first. ML/RL can disagree with the methodology, but it cannot trade or publish authority by itself. ## The methodology must explain before and after Pre-trade explanation answers why review, why paper intent or veto, which signal mattered, which risk capped it, and whether the state is paper-only. Post-trade explanation answers what happened, whether the outcome label is final, whether costs contaminated the result, and what the system learned. This is the user-facing difference between a score and a methodology. A score asks for trust. An explanation earns it. ## FAQ ### Is Orrery's methodology a trading strategy? It is a source-backed interpretation and paper-validation methodology. Public Orrery is read-only and research-only; live execution requires separate explicit operator approval. ### Why are vetoes important? Because avoiding unclear, illiquid, stale, or manipulation-prone markets can be as valuable as finding attractive markets. ### Can ML improve the methodology? Yes, but only as shadow evaluation first. ML can propose alternatives and be compared with OPE/calibration/drift checks before any authority is considered. ## Related Orrery resources - [Public methodology](https://orrery.me/methodology) - [Methodology evidence card](https://orrery.me/opportunities?methodology=evidence) - [Trust](https://orrery.me/trust) - [Receipts](https://orrery.me/receipts) - [Shadow agent docs](https://orrery.me/docs/agents/decision-api) Orrery is not affiliated with Polymarket and does not provide investment, legal, or tax advice.