Prediction Market
Glossary

A contract that pays a fixed amount — typically $1.00 — if a specific event occurs, and $0 if it does not. The price of a binary contract, expressed as a decimal between 0 and 1, is directly interpretable as a probability. Most prediction market contracts are binary options.

A scoring rule that measures the accuracy of probabilistic forecasts. Calculated as the mean squared error between stated probabilities and actual binary outcomes: BS = (1/n) Σ (f_t − o_t)², where f_t is the forecast probability and o_t is the outcome (1 or 0). Scores range from 0 (perfect) to 1 (perfectly wrong). Lower is better. A naïve baseline of always predicting 0.5 scores 0.25.

A measure of how well a forecaster's stated probabilities match observed outcome frequencies over many predictions. A perfectly calibrated forecaster who assigns "70% probability" to a class of events is correct on exactly 70% of those events. Calibration can be assessed visually using a reliability diagram, or numerically. Prediction markets tend to be better calibrated than individual experts because financial incentives penalize overconfidence.

A standardized identifier used to match the same event across different prediction market platforms, which often use different naming conventions, date formats, and team abbreviations. Without a canonical key, aggregating data from multiple sources produces duplicates and mismatches. A common format is sport:team1-team2:date.

A qualitative or quantitative label indicating how reliable a consensus probability estimate is, based on the number of contributing data sources and the degree of agreement between them. Common ratings: LOW (few sources or high spread), MEDIUM (moderate agreement), HIGH (multiple sources, tight spread). Confidence ratings help consumers of consensus data weight estimates appropriately.

An aggregated probability estimate derived by combining data from multiple independent prediction markets into a single figure per event. Unlike a raw price from one platform, a consensus probability pools information from multiple participant bases, reducing the impact of any one platform's idiosyncrasies, liquidity constraints, or user-base biases. Consensus probabilities are typically updated continuously as underlying market prices change.

A condition where different prediction markets assign materially different probabilities to the same event. High divergence — typically 8+ percentage points between platforms — indicates that information is being incorporated unevenly across markets, or that liquidity differences are creating pricing gaps. Divergence often decreases rapidly as new information enters the market and prices converge.

The theory that asset prices fully reflect all available information, making it impossible to consistently achieve above-average returns through analysis alone. Applied to prediction markets: if EMH holds, market prices are the best possible probability estimates at any given moment. In practice, prediction markets are believed to be more efficient than polls but less efficient than deep financial markets due to liquidity constraints and participant composition.

A financial contract whose value is determined by the outcome of a specific real-world event, such as an election result, a sporting outcome, or an economic announcement. Event contracts are the primary instrument traded on regulated prediction market platforms. They differ from options in that payoffs are binary and predetermined.

The price at which a contract should be priced given the available information — a model-derived probability estimate independent of current market prices. Fair value calculations use external data sources (polling averages, performance statistics, economic models) to produce an expected probability. The difference between a contract's market price and its fair value is sometimes called the "edge."

The probability of an outcome as directly encoded in the current price of a binary prediction market contract. Since binary contracts pay $1 on a win and $0 on a loss, the contract price equals the market's collective probability estimate. No conversion is needed: a price of $0.63 implies a 63% probability. This is simpler than converting from traditional pricing, which include a platform margin.

The process by which markets combine dispersed private information held by many different participants into a single consensus price. Each participant knows something different — local conditions, domain expertise, early data signals. Market incentives cause them to reveal this information through their trades, with prices adjusting until all available information is incorporated. This is the primary mechanism explaining why prediction markets often show strong calibration relative to individual expert forecasts.

The total volume of contracts available in a prediction market without significantly moving the price. High liquidity markets can absorb large orders with minimal price impact, making their prices more reliable. Low liquidity markets can be easily distorted by a single large order. Liquidity varies significantly across events: presidential elections typically have far more liquidity than local races or niche sporting events.

An entity (person, firm, or automated system) that continuously quotes both buy and sell prices for a contract, profiting from the bid-ask spread while providing liquidity to other participants. In traditional prediction markets, market makers may be human or algorithmic. In decentralized prediction markets, automated market maker (AMM) mechanisms set prices algorithmically based on the relative volume of yes vs. no contracts.

A data point or pattern derived from prediction market prices that carries informational value about the likely outcome of a future event. Market signals include current probability levels, probability changes over time, divergence between platforms, and convergence patterns following news events. Market signals are used for informational purposes — to contextualize news, assess uncertainty, or identify events with high information value.

The change in a contract's implied probability over a specified time interval. Movement is typically expressed in percentage points (e.g., "+4pp") or as a percentage of the previous price. Tracking movement highlights events where market opinion is shifting — either due to new information, changing conditions, or convergence with other markets. Direction (up or down) indicates whether the event is becoming more or less likely according to the market.

The number of independent prediction market platforms contributing data to a consensus probability estimate for a given event. Higher n-platform counts produce more reliable consensus estimates, since each additional independent source reduces the influence of any single platform's biases or liquidity constraints. Events with n=1 should be treated as single-source estimates, not true consensus.

A real-time list of all outstanding buy and sell orders for a contract, organized by price level. The order book shows market depth — how much volume is available at each price — and determines the best available price for any given order size. Prediction markets with deep order books (many resting orders at multiple price levels) are more liquid and their prices are more reliable indicators of true consensus probability.

A market where participants hold contracts whose value depends on the outcome of future events. Prices emerge from the interaction of buyers and sellers — each expressing their probability beliefs through their willingness to pay — and aggregate into a consensus probability estimate. Research consistently shows prediction markets are better calibrated than polls, individual expert forecasts, and most structured forecasting processes. Regulated prediction markets in the US operate under CFTC oversight.

The process by which prices on different prediction market platforms move toward each other over time. When the same event trades at significantly different prices across platforms, divergence analysisurs have an incentive to buy on the cheaper platform and sell on the more expensive one. This activity compresses the spread until prices are near-equal or the transaction costs make further divergence analysis uneconomical. High-divergence events often exhibit rapid convergence following major news.

A mathematical function that assigns probabilities to all possible outcomes of an uncertain event. For binary prediction market contracts (yes/no), the distribution is a Bernoulli distribution parameterized by a single probability p. For multi-outcome markets (e.g., "Which party prevails in the election?"), the distribution spans all candidates and must sum to 100%. Comparing distributions across platforms reveals where informational differences are most pronounced.

A prediction market operating under regulatory oversight, such as approval from the US CFTC (Commodity Futures Commission) as a Designated Contract Market (DCM). Regulated platforms must meet standards for market integrity, participant protection, and contract design. Regulatory compliance provides important consumer protections and ensures that market prices reflect genuine participant beliefs rather than manipulation.

The specific, predetermined rules that determine how a prediction market contract will be settled. Clear resolution criteria are essential for market integrity: participants need to know exactly what conditions trigger a "Yes" payout before they participate. Ambiguous resolution criteria can lead to disputes and erode trust. Well-designed criteria specify the data source, time window, and any edge-case handling.

The process by which a prediction market resolves — the real-world outcome is determined against the resolution criteria, and contracts are paid out accordingly. "Yes" holders receive $1 per contract; "No" holders receive $0 per contract, or vice versa depending on the outcome. Settlement may be immediate (for events with clear outcomes) or delayed (waiting for official data releases). Historical settlement data enables calibration analysis.

The difference between the expected price of a trade and the actual execution price, caused by insufficient liquidity at the target price level. In thin prediction markets, placing a large order exhausts available contracts at the best price and requires buying at progressively worse prices. Slippage is a key consideration for high-volume participants — it effectively increases the cost of entering or exiting positions in illiquid markets.

A small, high-density line chart — typically without axes or labels — that shows the trend of a data series over time in a compact space. In prediction market data, sparklines show the recent probability history for an event, enabling quick visual assessment of trend direction and volatility without requiring a full chart interface. Sparklines are commonly encoded as small arrays of historical values in APIs.

The difference in implied probability between the highest and lowest estimates for the same event across different prediction market platforms. A spread of 5 percentage points or less generally indicates good consensus; spreads above 10 points suggest meaningful disagreement between markets, which may reflect information asymmetry, liquidity differences, or regulatory constraints affecting one platform's participants. Spread is a key input to confidence rating calculations.

A prediction market with low market volume and limited order book depth, where even small trades can significantly affect the price. Thin markets are more susceptible to noise, manipulation, and stale prices — the last trade may have occurred hours ago and no longer reflects current information. Prices in thin markets are less reliable probability estimates than prices in deep, active markets.

A sequence of data points indexed by time, capturing how a value changes over regular intervals. In prediction markets, time-series data captures the evolution of implied probability for each event from market open to settlement. Time-series data enables trend analysis, volatility measurement, calibration studies, and backtesting of forecasting models. Each snapshot typically records the contract price, volume, and timestamp.

The empirical observation that the aggregate judgment of a large, diverse, and independent group is often more accurate than any individual expert opinion. Conditions for wisdom of crowds to hold: diversity of opinion, independence of participants, decentralization, and an aggregation mechanism. Prediction markets are specifically designed to capture and aggregate crowd wisdom via price discovery. The classic example is Galton's observation that the median crowd guess for an ox's weight (1,207 lbs) was within 1% of the actual weight (1,198 lbs).

An average in which each data point is multiplied by a weight reflecting its relative importance before being summed and divided by the total weight. In prediction market aggregation, weights may be assigned based on each platform's liquidity, historical calibration accuracy, or n_participants. A liquidity-weighted average gives more influence to high-volume platforms, which are presumed to be more efficient. A naive (equal-weight) average treats all platforms identically regardless of volume.