Appendix E: Glossary

This glossary defines more than 200 key terms used throughout Learning Prediction Markets — From Concepts to Strategies. Each entry indicates the chapter where the term is first introduced or most extensively discussed. Cross-references are provided to help readers navigate related concepts.

Accuracy (Chapter 9) : The proportion of predictions that are correct when evaluated against realized outcomes. In prediction markets, accuracy is often less informative than calibration or proper scoring because it ignores the confidence level of the prediction. See also: calibration, Brier score.

Adverse selection (Chapter 14) : A situation in which a market maker systematically trades against better-informed counterparties, leading to losses. In prediction markets, adverse selection arises when informed traders exploit stale prices before the market maker can update. See also: Glosten-Milgrom model, market maker, information aggregation.

Alpha (Chapter 30) : The excess return of a trading strategy relative to a benchmark, after adjusting for risk. In prediction markets, alpha represents the portion of profit attributable to genuine forecasting skill rather than luck or market-wide movements. See also: edge, Sharpe ratio.

AMM (Chapter 5) : See automated market maker.

Anchoring (Chapter 11) : A cognitive bias in which individuals rely too heavily on an initial piece of information (the "anchor") when making subsequent judgments. In prediction markets, traders may anchor on the current market price rather than performing independent analysis. See also: behavioral bias, availability heuristic.

API (Chapter 19) : Application Programming Interface. A set of protocols and tools allowing software programs to communicate with prediction market platforms, enabling automated trading, data retrieval, and order management. See also: web3.py, REST API.

Arbitrage (Chapter 6) : The simultaneous purchase and sale of equivalent or related contracts across different markets or within the same market to profit from price discrepancies without taking on net risk. See also: cross-platform arbitrage, temporal arbitrage, Dutch book.

ARIMA (Chapter 24) : AutoRegressive Integrated Moving Average. A class of time-series models that captures temporal dependencies in data by combining autoregressive terms, differencing for stationarity, and moving average terms. Used in prediction markets to model the evolution of contract prices over time. See also: stationarity, backtest.

Ask price (Chapter 4) : The lowest price at which a seller is willing to sell a contract. The ask price, together with the bid price, defines the bid-ask spread. See also: bid price, bid-ask spread, order book.

Automated market maker (AMM) (Chapter 5) : An algorithm that provides continuous liquidity by quoting buy and sell prices for prediction market contracts according to a mathematical pricing function. AMMs eliminate the need for a traditional order book by pooling funds and adjusting prices based on trading activity. See also: LMSR, CPMM, liquidity.

Availability heuristic (Chapter 11) : A cognitive shortcut in which people estimate the probability of an event based on how easily examples come to mind, rather than on objective frequency data. Dramatic or recent events tend to be overweighted. See also: anchoring, behavioral bias.

Backtesting (Chapter 25) : The process of evaluating a trading strategy or predictive model using historical data to estimate how it would have performed in the past. Proper backtesting requires guarding against look-ahead bias and overfitting. See also: walk-forward validation, overfitting.

Base rate (Chapter 8) : The unconditional or prior probability of an event, before incorporating any specific evidence. Base rate neglect, a common forecasting error, involves ignoring this background frequency. See also: Bayesian updating, prior.

Bayesian updating (Chapter 8) : The process of revising a probability estimate in light of new evidence using Bayes' theorem. In prediction markets, Bayesian updating describes how a rational trader should adjust their beliefs about an outcome when new information arrives. See also: prior, posterior, likelihood.

Behavioral bias (Chapter 11) : A systematic deviation from rational judgment caused by cognitive or emotional factors. Common biases affecting prediction market participants include anchoring, overconfidence, the disposition effect, and herding. See also: anchoring, overconfidence, prospect theory.

Beta distribution (Chapter 8) : A continuous probability distribution defined on the interval [0, 1], parameterized by two shape parameters. Commonly used as a conjugate prior for Bernoulli and binomial likelihoods, making it a natural choice for modeling beliefs about the probability of a binary event. See also: Bayesian updating, prior.

Bid-ask spread (Chapter 4) : The difference between the highest bid price and the lowest ask price in a market. The spread represents a transaction cost for traders and a source of revenue for market makers. Narrower spreads indicate higher liquidity. See also: bid price, ask price, liquidity, market maker.

Bid price (Chapter 4) : The highest price at which a buyer is willing to purchase a contract. See also: ask price, bid-ask spread.

Binary contract (Chapter 2) : A prediction market contract that pays a fixed amount (typically $1 or 1 unit) if a specified event occurs and nothing otherwise. The market price of a binary contract is commonly interpreted as the market's implied probability of the event. See also: event contract, outcome token.

Blockchain (Chapter 16) : A distributed, append-only ledger that records transactions in blocks linked by cryptographic hashes. Blockchains provide the settlement infrastructure for decentralized prediction markets, ensuring transparency, immutability, and censorship resistance. See also: Ethereum, smart contract, consensus mechanism.

Bollinger band (Chapter 24) : A technical analysis indicator consisting of a moving average plus and minus a fixed number of standard deviations. In prediction market contexts, Bollinger bands can help identify periods when a contract price has deviated unusually far from its recent mean. See also: mean reversion, volatility.

Brier score (Chapter 9) : A proper scoring rule that measures the accuracy of probabilistic predictions by computing the mean squared error between predicted probabilities and binary outcomes. Lower Brier scores indicate better calibration and discrimination. Defined as BS = (1/N) sum(f_i - o_i)^2. See also: proper scoring rule, calibration, logarithmic score.

Bundle trading (Chapter 6) : The simultaneous purchase or sale of a complete set of outcome contracts in a multi-outcome market. Because a complete bundle always pays out the fixed total, bundle trading enables arbitrage when the sum of individual contract prices deviates from the payout amount. See also: arbitrage, multi-outcome contract.

Calibration (Chapter 9) : The property that, among all events assigned a predicted probability of p, approximately a fraction p actually occur. A perfectly calibrated forecaster's reliability diagram falls along the 45-degree line. See also: reliability diagram, Brier score, overconfidence, underconfidence.

CFTC (Chapter 38) : Commodity Futures Trading Commission. The U.S. federal agency responsible for regulating futures and options markets. The CFTC has jurisdiction over event contracts traded on designated contract markets and has issued no-action letters to certain prediction market platforms. See also: DCM, no-action letter, Kalshi.

Chainlink (Chapter 17) : A decentralized oracle network that provides external data feeds to smart contracts on various blockchains. In decentralized prediction markets, Chainlink oracles can supply the real-world outcome data needed for contract resolution. See also: oracle, smart contract.

CLOB (Chapter 4) : Central Limit Order Book. A trading system that matches buy and sell orders based on price-time priority. Traditional prediction market exchanges such as Kalshi and PredictIt use CLOBs. See also: order book, price-time priority, matching engine.

Combinatorial market (Chapter 7) : A prediction market that allows trading on combinations of outcomes across multiple related questions. Combinatorial markets can express complex conditional beliefs but face computational challenges in maintaining consistent prices across exponentially many combinations. See also: conditional market, LMSR.

Conditional market (Chapter 7) : A prediction market in which contracts pay off contingent on both a specified event and a conditioning event. For example, "What will GDP growth be, given that Policy X is enacted?" Conditional markets are foundational to decision markets and futarchy. See also: decision market, futarchy, combinatorial market.

Conditional token framework (Chapter 17) : A smart contract standard developed by Gnosis for creating and trading conditional outcome tokens. The framework supports combinatorial markets by allowing tokens to represent positions contingent on multiple conditions. See also: Gnosis, outcome token, ERC-20.

Confirmation bias (Chapter 11) : The tendency to search for, interpret, and recall information in a way that confirms one's preexisting beliefs. In prediction markets, confirmation bias can cause traders to overweight evidence supporting their current position. See also: behavioral bias, overconfidence.

Consensus mechanism (Chapter 16) : The protocol by which nodes in a distributed network agree on the current state of the blockchain. Common mechanisms include proof-of-work and proof-of-stake. The choice of consensus mechanism affects the speed, cost, and security of decentralized prediction markets. See also: blockchain, Ethereum, gas.

Continuous double auction (Chapter 4) : A trading mechanism in which buyers and sellers can submit orders at any time, and trades execute whenever a buy order's price meets or exceeds a sell order's price. Most order-book-based prediction markets operate as continuous double auctions. See also: CLOB, matching engine, price-time priority.

Contrarian (Chapter 30) : A trading strategy or philosophy that involves taking positions opposite to the prevailing market sentiment. In prediction markets, contrarian strategies can be profitable when crowd sentiment overshoots due to herding or recency bias. See also: herding, mean reversion, behavioral bias.

Cost function market maker (Chapter 5) : A class of automated market makers defined by a convex cost function C(q) that determines the total payment required to move the market to a given state vector q. The LMSR is the most widely studied example. See also: LMSR, AMM, subsidy.

CPMM (Chapter 5) : Constant Product Market Maker. An AMM design where the product of token reserves remains constant (x * y = k). Originally popularized by Uniswap for token swaps, CPMMs have been adapted for prediction markets though they have different liquidity properties than LMSR. See also: AMM, LMSR, impermanent loss.

Cross-platform arbitrage (Chapter 31) : The practice of exploiting price differences for equivalent contracts listed on different prediction market platforms. For example, buying a contract priced at $0.40 on Platform A and selling the equivalent at $0.45 on Platform B. See also: arbitrage, temporal arbitrage.

Data pipeline (Chapter 20) : A sequence of automated steps that collects, cleans, transforms, and delivers data for use in prediction market models and trading systems. Robust data pipelines are essential for real-time strategy execution. See also: feature store, MLOps.

DCM (Chapter 38) : Designated Contract Market. A regulatory classification under U.S. law for exchanges authorized by the CFTC to list and trade futures and options, including event contracts. Kalshi operates as a DCM. See also: CFTC, Kalshi, regulation.

Decision market (Chapter 40) : A prediction market designed to inform a specific decision by estimating the consequences of alternative actions. Decision markets condition outcome predictions on which action is chosen, enabling decision-makers to compare expected outcomes across options. See also: conditional market, futarchy.

DeFi (Chapter 16) : Decentralized Finance. A broad category of financial services built on blockchain infrastructure that operate without traditional intermediaries. Decentralized prediction markets are a subset of DeFi applications. See also: blockchain, smart contract, Ethereum.

Depth chart (Chapter 4) : A graphical representation of the cumulative buy and sell orders at various price levels in an order book. Depth charts help traders visualize the available liquidity and potential price impact of large orders. See also: order book, liquidity, market impact.

Differential privacy (Chapter 37) : A mathematical framework for quantifying and limiting the privacy risk associated with data analysis. In prediction market contexts, differential privacy can protect sensitive information about individual participants while still allowing meaningful aggregate analysis. See also: privacy, regulation.

Discrimination (Chapter 9) : The ability of a forecasting system to distinguish between events that will occur and events that will not. A model with good discrimination assigns higher probabilities to events that actually happen. See also: calibration, Brier score, sharpness.

Disposition effect (Chapter 11) : A behavioral bias in which traders are inclined to sell winning positions too early (to lock in gains) and hold losing positions too long (to avoid realizing losses). Rooted in prospect theory's asymmetric treatment of gains and losses. See also: prospect theory, behavioral bias.

Diversification (Chapter 29) : The practice of spreading capital across multiple uncorrelated or weakly correlated prediction market positions to reduce overall portfolio risk. Diversification lowers the variance of returns without necessarily reducing expected return. See also: portfolio Kelly, risk of ruin, correlation.

Drawdown (Chapter 29) : The decline in portfolio value from a peak to a subsequent trough, usually expressed as a percentage. Maximum drawdown is a key risk metric for prediction market traders. See also: maximum drawdown, risk of ruin, Sharpe ratio.

Dutch book (Chapter 3) : A set of bets that guarantees a loss for the bettor regardless of the outcome, arising from incoherent (non-additive) probability assignments. The Dutch book theorem shows that only probability distributions satisfying the axioms of probability are immune to Dutch books. See also: arbitrage, coherence.

ECE (Chapter 10) : Expected Calibration Error. A scalar summary of calibration computed by binning predictions, calculating the absolute difference between average predicted probability and observed frequency within each bin, and taking a weighted average. See also: calibration, reliability diagram.

Edge (Chapter 30) : The expected profit per unit wagered on a prediction market position, representing the difference between a trader's assessed probability and the market price (adjusted for transaction costs). A positive edge means the trader believes the position is mispriced in their favor. See also: expected value, alpha.

Ensemble (Chapter 23) : A modeling technique that combines predictions from multiple individual models to produce a more accurate and robust forecast. Common ensemble methods include bagging, boosting, and stacking. See also: stacking, random forest, XGBoost.

ERC-20 (Chapter 16) : A widely adopted token standard on the Ethereum blockchain that defines a common interface for fungible tokens. Prediction market outcome tokens are often implemented as ERC-20 tokens. See also: Ethereum, outcome token, conditional token framework.

Ethereum (Chapter 16) : A decentralized blockchain platform that supports smart contracts and decentralized applications. Ethereum serves as the primary infrastructure for many decentralized prediction markets, including Augur, Gnosis, and Polymarket. See also: blockchain, smart contract, EVM, Solidity.

EVM (Chapter 16) : Ethereum Virtual Machine. The runtime environment for executing smart contracts on Ethereum and compatible blockchains. The EVM processes the bytecode compiled from languages like Solidity. See also: Ethereum, smart contract, Solidity.

Event contract (Chapter 2) : A financial instrument whose payoff depends on the occurrence or non-occurrence of a specified real-world event. Event contracts are the fundamental building blocks of prediction markets. See also: binary contract, scalar contract, multi-outcome contract.

Event sourcing (Chapter 20) : A software architecture pattern in which state changes are stored as an immutable sequence of events rather than by overwriting the current state. Event sourcing is useful for building auditable, replayable prediction market trading systems. See also: data pipeline.

Expected value (Chapter 3) : The probability-weighted average of all possible payoffs from a bet or position. A prediction market trade has positive expected value when the trader's estimated probability exceeds the price paid (for a Yes contract) or falls below it (for a No contract). See also: edge, Kelly criterion.

Extremizing (Chapter 13) : The transformation of aggregated probability forecasts by pushing them further from 0.5 to correct for informational redundancy among forecasters. Extremizing addresses the tendency of simple averages of probabilities to be insufficiently confident. See also: calibration, recalibration.

Favorite-longshot bias (Chapter 12) : The empirical regularity that favorites (high-probability outcomes) tend to be underpriced and longshots (low-probability outcomes) tend to be overpriced in betting and prediction markets. See also: overround, behavioral bias, calibration.

Feature engineering (Chapter 22) : The process of selecting, transforming, and creating input variables for a predictive model. In prediction markets, features might include historical price data, polling numbers, economic indicators, and sentiment scores. See also: feature store, random forest, XGBoost.

Feature store (Chapter 22) : A centralized repository for managing, versioning, and serving precomputed features for machine learning models. Feature stores help ensure consistency between training and production environments in prediction market ML systems. See also: MLOps, data pipeline.

Flash loan (Chapter 18) : An uncollateralized loan on a blockchain that must be borrowed and repaid within a single transaction. Flash loans can be used for arbitrage on decentralized prediction markets but also create attack vectors. See also: DeFi, MEV, arbitrage.

Fractional Kelly (Chapter 28) : A conservative variant of the Kelly criterion that bets a fixed fraction (e.g., half or quarter) of the full Kelly stake. Fractional Kelly sacrifices some expected growth rate in exchange for substantially reduced variance and drawdown risk. See also: Kelly criterion, portfolio Kelly, risk of ruin.

Front-running (Chapter 18) : The practice of placing orders ahead of known pending transactions to profit from the anticipated price impact. In decentralized prediction markets, front-running can occur via MEV extraction on the blockchain. See also: MEV, adverse selection.

Futarchy (Chapter 40) : A governance system proposed by Robin Hanson in which prediction markets are used to evaluate and select policies. Under futarchy, society would "vote on values, but bet on beliefs," using conditional prediction markets to determine which policies best achieve agreed-upon objectives. See also: decision market, conditional market, Hanson.

Gas (Chapter 16) : The unit of computational effort required to execute operations on the Ethereum blockchain. Gas costs represent a significant transaction cost for participants in decentralized prediction markets. See also: Ethereum, EVM, transaction cost.

Glosten-Milgrom model (Chapter 14) : A theoretical model of market microstructure in which a market maker sets bid and ask prices knowing that some fraction of incoming traders are informed. The model shows how the bid-ask spread reflects the degree of adverse selection. See also: adverse selection, bid-ask spread, market maker.

Gnosis (Chapter 17) : A blockchain-based platform for creating and trading prediction market contracts. Gnosis developed the conditional token framework and has been a pioneer in decentralized prediction market infrastructure. See also: conditional token framework, Ethereum, DeFi.

Governance (Chapter 40) : The systems and processes by which organizations or communities make collective decisions. Prediction markets can serve as governance tools by aggregating dispersed information to inform decision-making. See also: futarchy, decision market.

Governance token (Chapter 17) : A cryptocurrency token that grants holders voting rights in the governance of a decentralized protocol. Some prediction market platforms issue governance tokens to align incentives between the platform and its users. See also: DeFi, REP token.

Herding (Chapter 12) : The tendency of individuals to mimic the actions of a larger group, regardless of their own private information. In prediction markets, herding can cause prices to deviate from fundamental values and reduce information aggregation efficiency. See also: information cascade, behavioral bias.

Howey test (Chapter 38) : A legal test established by the U.S. Supreme Court to determine whether a transaction qualifies as an "investment contract" (and therefore a security) under federal securities law. The Howey test is relevant for determining the regulatory classification of prediction market tokens. See also: regulation, SEC.

Hyperparameter (Chapter 23) : A configuration setting for a machine learning model that is specified before training begins, rather than learned from data. Examples include learning rate, regularization strength, and tree depth. Hyperparameter tuning is critical for building effective prediction market models. See also: random forest, XGBoost, cross-validation.

Impermanent loss (Chapter 18) : The reduction in value that a liquidity provider experiences in an AMM compared to simply holding the underlying tokens, caused by divergence in token prices. In prediction markets, impermanent loss is a cost of subsidizing market liquidity. See also: AMM, CPMM, liquidity mining.

Implied probability (Chapter 2) : The probability of an event as inferred from the market price of the corresponding prediction market contract. For a binary contract paying $1 on the event, the implied probability approximately equals the contract price (before adjusting for transaction costs and overround). See also: binary contract, overround.

Information aggregation (Chapter 1) : The process by which a prediction market combines the dispersed knowledge, beliefs, and information of its participants into a single consensus price. This is the primary function and theoretical justification for prediction markets. See also: wisdom of crowds, efficient market hypothesis.

Information cascade (Chapter 12) : A phenomenon in which individuals sequentially observe the actions of predecessors and rationally choose to follow them rather than act on their own private information, potentially leading to incorrect consensus. See also: herding, behavioral bias.

Insider trading (Chapter 39) : Trading in a prediction market based on material non-public information about the event outcome. The legal and ethical treatment of insider trading in prediction markets differs from traditional securities markets and remains a topic of debate. See also: adverse selection, regulation.

Isotonic regression (Chapter 10) : A non-parametric method for recalibrating probability forecasts by fitting a monotonically non-decreasing function to map raw model outputs to calibrated probabilities. Isotonic regression is more flexible than Platt scaling but requires more data. See also: calibration, Platt scaling, recalibration.

Kalshi (Chapter 38) : A CFTC-regulated prediction market exchange operating as a Designated Contract Market in the United States. Kalshi offers event contracts on a variety of topics including economics, weather, and politics. See also: CFTC, DCM, event contract.

Kelly criterion (Chapter 28) : A formula for optimal bet sizing that maximizes the expected logarithmic growth rate of wealth. For a binary bet with edge e and odds b, the Kelly fraction is f* = e/b. The Kelly criterion balances aggressive growth against the risk of large drawdowns. See also: fractional Kelly, portfolio Kelly, risk of ruin.

Kleros (Chapter 17) : A decentralized dispute resolution protocol that uses game-theoretic incentives and crowdsourced jurors to adjudicate disputes. Kleros can serve as an arbitration mechanism for decentralized prediction market resolution disputes. See also: Schelling point, oracle, smart contract.

KYC/AML (Chapter 38) : Know Your Customer / Anti-Money Laundering. Regulatory requirements that oblige financial service providers, including regulated prediction market platforms, to verify the identity of their customers and monitor for suspicious activity. See also: CFTC, regulation.

Layer 2 (Chapter 18) : A scaling solution built on top of a base blockchain (Layer 1) that processes transactions off the main chain to achieve higher throughput and lower costs. Layer 2 solutions such as rollups are increasingly used by decentralized prediction markets to reduce gas fees. See also: Ethereum, gas, blockchain.

Likelihood (Chapter 8) : In Bayesian inference, the probability of observing the evidence given a particular hypothesis. The likelihood function connects observed data to model parameters and is combined with the prior via Bayes' theorem to produce the posterior. See also: Bayesian updating, prior, posterior.

Limit order (Chapter 4) : An order to buy or sell a prediction market contract at a specified price or better. Limit orders rest in the order book until matched or cancelled and provide liquidity to the market. See also: market order, order book, CLOB.

Liquidity (Chapter 4) : The ease with which contracts can be bought or sold without causing significant price movement. High liquidity, characterized by narrow bid-ask spreads and deep order books, reduces transaction costs for traders. See also: bid-ask spread, market impact, depth chart.

Liquidity mining (Chapter 18) : A DeFi mechanism in which participants earn token rewards for providing liquidity to an AMM pool. Some decentralized prediction markets use liquidity mining to incentivize market makers and ensure sufficient trading depth. See also: AMM, impermanent loss, DeFi.

LMSR (Chapter 5) : Logarithmic Market Scoring Rule. An automated market maker designed by Robin Hanson that uses a logarithmic cost function to price contracts. The LMSR provides bounded loss for the market operator (subsidizer) and has desirable information-theoretic properties. See also: AMM, cost function market maker, LS-LMSR, subsidy.

Logarithmic score (Chapter 9) : A proper scoring rule that evaluates a probabilistic forecast by taking the logarithm of the probability assigned to the event that actually occurred. The logarithmic score is strictly proper and penalizes confident but wrong predictions severely. See also: proper scoring rule, Brier score.

Log-likelihood (Chapter 8) : The logarithm of the likelihood function, used for computational convenience in optimization and Bayesian analysis. Maximizing the log-likelihood is equivalent to maximizing the likelihood. See also: likelihood, Bayesian updating.

LS-LMSR (Chapter 5) : Liquidity-Sensitive Logarithmic Market Scoring Rule. A variant of the LMSR in which the liquidity parameter adjusts dynamically based on trading volume, increasing liquidity as market interest grows. See also: LMSR, AMM, liquidity.

Manifold Markets (Chapter 1) : A play-money prediction market platform that allows users to create and trade on user-generated questions. Manifold Markets has been widely used for research, experimentation, and community forecasting. See also: play money, prediction market.

Margin (Chapter 29) : The collateral or funds that a trader must deposit to open and maintain a leveraged position. In prediction markets, margin requirements determine the capital efficiency of trading strategies. See also: position, risk of ruin.

Market impact (Chapter 14) : The effect of a trade on the prevailing market price. Large orders move prices more than small ones, creating a cost that must be accounted for in strategy design. See also: liquidity, slippage, VWAP.

Market maker (Chapter 5) : An entity that provides liquidity to a market by continuously quoting bid and ask prices, profiting from the spread while bearing inventory risk and adverse selection costs. See also: AMM, Glosten-Milgrom model, bid-ask spread, adverse selection.

Market order (Chapter 4) : An order to buy or sell a contract immediately at the best available price. Market orders guarantee execution but not price and "take" liquidity from the order book. See also: limit order, slippage.

Matching engine (Chapter 4) : The software system that pairs incoming buy and sell orders according to priority rules (typically price-time priority) and executes trades. See also: CLOB, continuous double auction, price-time priority.

Maximum drawdown (Chapter 29) : The largest peak-to-trough decline in portfolio value over a specified period. Maximum drawdown is a critical risk metric for evaluating prediction market trading strategies. See also: drawdown, risk of ruin, Sharpe ratio.

Mean reversion (Chapter 24) : The tendency of a variable to return toward its long-run average over time. In prediction markets, mean reversion may appear when prices temporarily overshoot due to behavioral biases and then correct. See also: Bollinger band, contrarian.

Metaculus (Chapter 1) : A community forecasting platform that uses proper scoring rules rather than a market mechanism. Metaculus has built one of the largest public track records of calibrated probabilistic predictions. See also: proper scoring rule, calibration, play money.

MEV (Chapter 18) : Maximal Extractable Value. The profit that blockchain validators or searchers can extract by reordering, inserting, or censoring transactions within a block. MEV creates front-running risks for participants in decentralized prediction markets. See also: front-running, blockchain, DeFi.

MLflow (Chapter 26) : An open-source platform for managing the machine learning lifecycle, including experiment tracking, model packaging, and deployment. MLflow is used in production prediction market systems to maintain reproducibility and model governance. See also: MLOps, model registry.

MLOps (Chapter 26) : Machine Learning Operations. The set of practices that combines machine learning, DevOps, and data engineering to deploy and maintain ML models in production reliably. See also: MLflow, feature store, data pipeline.

Model registry (Chapter 26) : A centralized repository for versioning, staging, and managing machine learning models throughout their lifecycle. Model registries are part of MLOps infrastructure for prediction market trading systems. See also: MLOps, MLflow.

Monte Carlo simulation (Chapter 29) : A computational technique that uses repeated random sampling to estimate the distribution of possible outcomes. In prediction markets, Monte Carlo methods are used for portfolio risk assessment, strategy evaluation, and confidence interval estimation. See also: Value at Risk, risk of ruin.

Moral hazard (Chapter 39) : The risk that a participant in a prediction market may have the ability and incentive to influence the outcome of the event being predicted, creating a conflict of interest. See also: insider trading, manipulation, regulation.

Multi-outcome contract (Chapter 2) : A prediction market instrument with three or more mutually exclusive and exhaustive outcomes. Examples include "Which party will win the next election?" with multiple candidates. See also: binary contract, combinatorial market, bundle trading.

No-action letter (Chapter 38) : A letter from a regulatory agency (such as the CFTC) stating that the agency will not recommend enforcement action against the recipient for a specified activity. PredictIt operated under a CFTC no-action letter. See also: CFTC, regulation, PredictIt.

Noise trader (Chapter 14) : A market participant who trades for reasons unrelated to information about the fundamental value of the contract, such as entertainment, hedging, or random speculation. Noise traders provide essential cover for informed traders and facilitate market operation. See also: adverse selection, information aggregation.

Oracle (Chapter 17) : A mechanism for delivering external real-world data to a blockchain smart contract. Prediction markets require oracles to determine event outcomes and trigger contract resolution. See also: Chainlink, smart contract, resolution.

Order book (Chapter 4) : A real-time list of all outstanding buy and sell orders for a given contract, organized by price level. The order book reveals the available liquidity at each price point. See also: CLOB, depth chart, bid-ask spread.

Outcome token (Chapter 16) : A digital token representing a claim on one specific outcome of a prediction market event. Holding an outcome token entitles the holder to a payout if that outcome occurs. See also: ERC-20, conditional token framework, binary contract.

Overconfidence (Chapter 11) : A cognitive bias in which individuals assign probabilities that are too extreme — too close to 0 or 1 — relative to their actual knowledge. Overconfident forecasters exhibit poor calibration. See also: calibration, underconfidence, behavioral bias.

Overfitting (Chapter 23) : Building a model that captures noise rather than genuine signal in the training data, resulting in strong in-sample performance but poor out-of-sample generalization. Overfitting is a central risk in developing prediction market models. See also: cross-validation, backtesting, regularization.

Overround (Chapter 6) : The amount by which the sum of implied probabilities across all outcomes exceeds 100%, representing the market maker's or bookmaker's built-in margin. Also called the "vig" or "juice." See also: vig, implied probability, favorite-longshot bias.

Platt scaling (Chapter 10) : A post-hoc calibration method that fits a logistic regression to map raw model scores to calibrated probabilities. Platt scaling is parametric and works well when the calibration curve is approximately sigmoidal. See also: isotonic regression, calibration, recalibration.

Play money (Chapter 1) : Virtual currency used in prediction markets where participants do not wager real money. Play-money markets can still produce informative forecasts, though they may suffer from weaker incentives. See also: Manifold Markets, Metaculus.

Polymarket (Chapter 17) : A decentralized prediction market platform built on Ethereum (and later Polygon) that allows users to trade on the outcomes of real-world events using cryptocurrency. Polymarket has become one of the largest decentralized prediction markets by volume. See also: DeFi, Ethereum, outcome token.

Portfolio Kelly (Chapter 29) : An extension of the Kelly criterion to multiple simultaneous bets, optimizing the allocation of capital across a portfolio of prediction market positions while accounting for correlations. See also: Kelly criterion, fractional Kelly, diversification.

Position (Chapter 4) : A trader's current holdings in a specific prediction market contract, measured in the number of contracts held. A "long" position profits if the outcome occurs; a "short" position profits if it does not. See also: binary contract, margin.

Posterior (Chapter 8) : The updated probability distribution for a hypothesis after incorporating observed evidence via Bayes' theorem. The posterior combines the prior and the likelihood. See also: Bayesian updating, prior, likelihood.

PredictIt (Chapter 38) : A prediction market platform operated by Victoria University of Wellington under a CFTC no-action letter. PredictIt imposed position and contract limits and was a prominent U.S. real-money prediction market until regulatory changes. See also: CFTC, no-action letter, regulation.

Price-time priority (Chapter 4) : The order-matching rule used by most exchanges in which orders at the best price are filled first, with ties broken by the time of submission (earlier orders have priority). See also: CLOB, matching engine, continuous double auction.

Prior (Chapter 8) : The probability distribution representing beliefs about a parameter or hypothesis before observing new data. In Bayesian updating, the prior is combined with the likelihood to produce the posterior. See also: Bayesian updating, posterior, beta distribution.

Proper scoring rule (Chapter 9) : A scoring function for probabilistic forecasts that is maximized (in expectation) when the forecaster reports their true believed probabilities. The Brier score and logarithmic score are both strictly proper. See also: Brier score, logarithmic score, spherical score.

Prospect theory (Chapter 11) : A behavioral economic theory developed by Kahneman and Tversky describing how people evaluate risky decisions. Prospect theory features loss aversion, probability weighting, and reference dependence. See also: disposition effect, behavioral bias, loss aversion.

Random forest (Chapter 23) : An ensemble machine learning method that builds many decision trees on bootstrapped subsets of the data and averages their predictions. Random forests are robust to overfitting and widely used for prediction market feature-based models. See also: ensemble, XGBoost, hyperparameter.

Rate limiting (Chapter 19) : The practice of restricting the number of API requests a client can make within a given time window. Prediction market APIs impose rate limits to prevent abuse and ensure system stability. See also: API, data pipeline.

Recalibration (Chapter 10) : The process of adjusting the output of a probabilistic model so that its predictions become better calibrated. Common recalibration methods include Platt scaling and isotonic regression. See also: calibration, Platt scaling, isotonic regression.

Regime detection (Chapter 24) : The identification of distinct market states or "regimes" characterized by different statistical properties (e.g., high vs. low volatility, trending vs. mean-reverting). Hidden Markov models are a common tool for regime detection. See also: stationarity, ARIMA, hidden Markov model.

Regularization (Chapter 23) : A technique for preventing overfitting by adding a penalty term to the model's objective function that discourages excessive complexity. Common forms include L1 (Lasso) and L2 (Ridge) regularization. See also: overfitting, hyperparameter.

Reliability diagram (Chapter 9) : A plot of observed outcome frequency against predicted probability, typically computed in bins. A perfectly calibrated model produces points along the diagonal. See also: calibration, ECE, Brier score.

REP token (Chapter 17) : The native token of the Augur prediction market protocol, used by reporters who stake REP to report event outcomes and earn rewards for honest reporting. See also: oracle, governance token, Augur.

Resolution (Chapter 2) : The process of determining the actual outcome of a prediction market event and settling contracts accordingly. Resolution may be handled by a trusted authority, a decentralized oracle, or a consensus mechanism. See also: oracle, Chainlink, Kleros.

REST API (Chapter 19) : Representational State Transfer API. A common architectural style for web APIs, used by many prediction market platforms for data retrieval and order management. See also: API, rate limiting.

Risk of ruin (Chapter 28) : The probability that a trader's bankroll will be depleted entirely, rendering them unable to continue trading. Risk of ruin analysis is critical for determining appropriate position sizes. See also: Kelly criterion, fractional Kelly, maximum drawdown.

Scalar contract (Chapter 2) : A prediction market contract whose payout is proportional to the realized value of a continuous variable within a specified range. For example, a contract paying proportionally to GDP growth between 0% and 5%. See also: binary contract, multi-outcome contract.

Schelling point (Chapter 17) : A focal solution that people converge on in the absence of explicit coordination, identified by Thomas Schelling. In decentralized prediction markets, Schelling points underpin dispute resolution mechanisms where reporters independently converge on the truth. See also: Kleros, oracle, consensus mechanism.

Scoring rule (Chapter 9) : A function that assigns a numerical score to a probabilistic forecast based on the forecast and the realized outcome. Scoring rules are used to evaluate and incentivize forecaster performance. See also: proper scoring rule, Brier score, logarithmic score.

SEC (Chapter 38) : Securities and Exchange Commission. The U.S. federal agency responsible for regulating securities markets. Some prediction market contracts may be classified as securities, bringing them under SEC jurisdiction. See also: Howey test, regulation, CFTC.

Sharpe ratio (Chapter 30) : The ratio of a strategy's excess return (over the risk-free rate) to its standard deviation. The Sharpe ratio measures risk-adjusted performance and allows comparison across prediction market strategies with different risk profiles. See also: alpha, maximum drawdown, volatility.

Sharpness (Chapter 9) : The degree to which probabilistic forecasts are concentrated (close to 0 or 1) rather than hedged (close to 0.5). Sharpness is desirable only in combination with good calibration. See also: calibration, Brier score, resolution (in decomposition).

Slippage (Chapter 4) : The difference between the expected execution price and the actual price at which a trade is filled, caused by insufficient liquidity or rapid price movement. Slippage is a transaction cost that can erode the profitability of prediction market strategies. See also: market impact, liquidity, market order.

Smart contract (Chapter 16) : A self-executing program stored on a blockchain that automatically enforces the terms of an agreement when predetermined conditions are met. Prediction market contracts on decentralized platforms are implemented as smart contracts. See also: blockchain, Ethereum, Solidity.

Solidity (Chapter 16) : The primary programming language for writing smart contracts on Ethereum and EVM-compatible blockchains. Prediction market developers use Solidity to implement contract logic, token standards, and resolution mechanisms. See also: Ethereum, EVM, smart contract.

Spherical score (Chapter 9) : A proper scoring rule that evaluates a probabilistic forecast by normalizing the probability vector and computing the probability assigned to the observed outcome divided by the Euclidean norm. See also: proper scoring rule, Brier score, logarithmic score.

Stacking (Chapter 23) : An ensemble technique in which the predictions of multiple base models are used as inputs to a meta-model that learns to combine them optimally. Stacking often outperforms simple averaging for prediction market forecasting. See also: ensemble, random forest, XGBoost.

Stationarity (Chapter 24) : The property that the statistical characteristics of a time series (mean, variance, autocorrelation) do not change over time. Many time-series models, including ARIMA, assume stationarity or require differencing to achieve it. See also: ARIMA, regime detection.

Subsidy (Chapter 5) : The initial funding provided by the market operator to an automated market maker to bootstrap liquidity. In the LMSR, the maximum loss (and thus required subsidy) is bounded and determined by the liquidity parameter. See also: LMSR, AMM, liquidity.

Temporal arbitrage (Chapter 31) : Exploiting price discrepancies that arise over time as information propagates at different speeds across markets or as markets react with delay to new data. See also: arbitrage, cross-platform arbitrage.

Tilt (Chapter 32) : A state of emotional frustration or often triggered by a string of losses, that impairs a trader's judgment and leads to impulsive, suboptimal decision-making. Borrowed from poker terminology. See also: behavioral bias, disposition effect, risk of ruin.

Transaction cost (Chapter 6) : Any cost incurred when buying or selling prediction market contracts, including bid-ask spreads, exchange fees, gas fees (on blockchains), and slippage. Transaction costs reduce the net profitability of trades and must be factored into any strategy. See also: bid-ask spread, gas, slippage, overround.

Transformer (Chapter 25) : A deep learning architecture based on self-attention mechanisms, originally developed for natural language processing. Transformers can be applied to prediction market modeling for processing textual data (news, social media) and sequential market data. See also: feature engineering, ensemble.

UMA (Chapter 17) : Universal Market Access. A decentralized protocol for creating synthetic assets and prediction markets with an optimistic oracle design for dispute resolution. See also: oracle, DeFi, smart contract.

Underconfidence (Chapter 11) : A calibration failure in which predicted probabilities are too moderate — closer to 0.5 than warranted by the forecaster's actual information. The opposite of overconfidence. See also: calibration, overconfidence, extremizing.

Value at Risk (VaR) (Chapter 29) : A statistical measure of the maximum expected loss over a given time horizon at a specified confidence level. VaR is used in prediction market portfolio management to quantify downside risk. See also: Monte Carlo simulation, maximum drawdown, risk of ruin.

Vig (Chapter 6) : Short for vigorish. The commission or margin embedded in prediction market prices, equivalent to the overround. The vig represents the cost of trading and the market operator's revenue. See also: overround, transaction cost.

Volatility (Chapter 24) : A statistical measure of the dispersion of price changes over time, typically quantified as the standard deviation of returns. Higher volatility implies greater uncertainty and wider price swings for prediction market contracts. See also: Bollinger band, Sharpe ratio.

VWAP (Chapter 31) : Volume-Weighted Average Price. The average price of a contract weighted by the volume traded at each price level over a given period. VWAP is used as a benchmark for evaluating execution quality. See also: market impact, slippage.

Walk-forward validation (Chapter 25) : A backtesting methodology in which a model is trained on a rolling window of historical data and tested on the immediately subsequent period, then the window advances forward. Walk-forward validation better simulates real-time forecasting conditions than static train-test splits. See also: backtesting, overfitting, cross-validation.

Wash trading (Chapter 39) : The illegal or unethical practice of simultaneously buying and selling the same contract to create the illusion of market activity and volume. Wash trading can distort liquidity metrics and price signals. See also: manipulation, regulation.

Web3.py (Chapter 19) : A Python library for interacting with the Ethereum blockchain, used to programmatically trade on decentralized prediction markets, query smart contract state, and submit transactions. See also: Ethereum, API, smart contract.

Wisdom of crowds (Chapter 1) : The principle, articulated by James Surowiecki, that the aggregate judgment of a diverse, independent group of individuals is often more accurate than the judgment of any single expert. Prediction markets are a mechanism for harnessing the wisdom of crowds. See also: information aggregation, Surowiecki.

XGBoost (Chapter 23) : Extreme Gradient Boosting. A highly optimized implementation of gradient-boosted decision trees, known for strong performance in structured data prediction tasks. XGBoost is widely used for building prediction market forecasting models. See also: random forest, ensemble, hyperparameter.

Zero-knowledge proof (Chapter 18) : A cryptographic protocol that allows one party to prove knowledge of a fact to another party without revealing the fact itself. Zero-knowledge proofs have potential applications in privacy-preserving prediction markets, enabling trading without revealing positions or strategies. See also: blockchain, differential privacy.

Additional terms:

Aggregation (Chapter 13) : The process of combining multiple individual forecasts into a single consensus forecast. Methods include simple averaging, weighted averaging, and extremized aggregation. See also: extremizing, wisdom of crowds.

Augur (Chapter 17) : One of the earliest decentralized prediction market protocols, built on Ethereum. Augur uses REP token staking to incentivize honest outcome reporting. See also: REP token, Ethereum, DeFi.

Bayes' theorem (Chapter 8) : The mathematical formula relating the posterior probability of a hypothesis to its prior probability and the likelihood of the observed evidence: P(H|E) = P(E|H) * P(H) / P(E). See also: Bayesian updating, prior, posterior.

Benchmark (Chapter 30) : A reference standard against which the performance of a trading strategy or forecasting model is measured. Common benchmarks include naive models (e.g., predict the base rate) and market consensus. See also: alpha, Sharpe ratio.

Binomial distribution (Chapter 8) : The discrete probability distribution of the number of successes in a fixed number of independent Bernoulli trials. Used for modeling the frequency of binary outcomes in prediction markets. See also: beta distribution, Bayesian updating.

Black-Scholes model (Chapter 15) : A mathematical model for pricing options that assumes log-normal price dynamics. While designed for financial options, the Black-Scholes framework provides conceptual parallels for pricing prediction market contracts with time-varying uncertainty. See also: volatility, option.

Bootstrap (Chapter 25) : A statistical resampling technique that generates multiple datasets by sampling with replacement from the original data. Bootstrapping is used to estimate confidence intervals and model uncertainty. See also: random forest, Monte Carlo simulation.

Buy order (Chapter 4) : An instruction to purchase a specified number of contracts at a given price. See also: limit order, market order, order book.

Central limit theorem (Chapter 8) : A fundamental theorem in probability stating that the sum or average of a large number of independent random variables is approximately normally distributed, regardless of the underlying distribution. See also: normal distribution.

Coherence (Chapter 3) : The property of a set of probability assignments that are internally consistent and not susceptible to a Dutch book. Coherent probabilities satisfy the axioms of probability theory. See also: Dutch book, expected value.

Correlation (Chapter 29) : A statistical measure of the linear relationship between two variables. Understanding correlations between prediction market positions is essential for portfolio construction and risk management. See also: diversification, portfolio Kelly.

Cross-validation (Chapter 23) : A model evaluation technique that partitions data into complementary subsets for training and testing, repeating the process across multiple folds to produce robust performance estimates. See also: overfitting, walk-forward validation.

Decay (Chapter 24) : The tendency for the informational value or accuracy of a prediction to diminish as the information ages. Time decay affects the relevance of features and the value of stale market data. See also: feature engineering, temporal arbitrage.

Efficient market hypothesis (Chapter 3) : The theory that market prices fully reflect all available information, implying that it is impossible to consistently achieve above-market returns without taking on additional risk. Prediction markets test the weak, semi-strong, and strong forms of this hypothesis. See also: information aggregation, arbitrage.

Expected log-wealth (Chapter 28) : The expected value of the logarithm of terminal wealth, which the Kelly criterion maximizes. Maximizing expected log-wealth leads to the highest geometric growth rate over time. See also: Kelly criterion.

Fee (Chapter 6) : A charge imposed by a prediction market platform for executing trades, withdrawing funds, or maintaining positions. Fees are a component of overall transaction costs. See also: transaction cost, overround.

Geometric growth rate (Chapter 28) : The compound rate at which capital grows over time. The Kelly criterion produces the highest geometric growth rate among all fixed-fraction betting strategies. See also: Kelly criterion, expected log-wealth.

Good judgment (Chapter 13) : The capacity to make well-calibrated probabilistic assessments. The Good Judgment Project demonstrated that trained forecasters (superforecasters) can substantially outperform both markets and experts. See also: calibration, superforecaster, Tetlock.

Hidden Markov model (Chapter 24) : A statistical model in which the system being modeled is assumed to transition among hidden states according to a Markov process, with observable emissions at each state. HMMs are used for regime detection in market data. See also: regime detection, stationarity.

Incentive compatibility (Chapter 9) : The property that a mechanism provides participants with incentives to behave truthfully. Proper scoring rules are incentive-compatible because they reward honest probability reporting. See also: proper scoring rule, mechanism design.

Informed trader (Chapter 14) : A market participant who possesses private information about the likely outcome of an event and trades to profit from this informational advantage. See also: adverse selection, noise trader, Glosten-Milgrom model.

Inventory risk (Chapter 14) : The financial risk that a market maker faces from holding an unbalanced portfolio of contracts. Large inventory positions expose the market maker to directional risk. See also: market maker, adverse selection.

Leverage (Chapter 29) : The use of borrowed funds or margin to increase the size of a position beyond what the trader's own capital would permit. Leverage amplifies both gains and losses. See also: margin, risk of ruin.

Liquidity parameter (Chapter 5) : In the LMSR, the parameter b that controls the trade-off between liquidity (how much prices move per trade) and the maximum loss of the market maker. A larger b provides more liquidity but requires a larger subsidy. See also: LMSR, subsidy.

Liquidity provider (Chapter 5) : An entity that supplies funds to an automated market maker pool, enabling trading by others. Liquidity providers earn fees but bear impermanent loss risk. See also: AMM, impermanent loss, liquidity mining.

Log odds (Chapter 8) : The logarithm of the odds ratio, log(p / (1 - p)). Log odds provide a convenient scale for Bayesian updating where updates are additive. See also: Bayesian updating, logistic regression.

Logistic regression (Chapter 22) : A statistical model for binary classification that models the log odds of the positive class as a linear function of input features. Logistic regression is a common baseline model for prediction market binary outcome forecasting. See also: Platt scaling, calibration.

Loss aversion (Chapter 11) : The tendency for the pain of a loss to be felt more intensely than the pleasure of an equivalent gain, a core element of prospect theory. See also: prospect theory, disposition effect.

Manipulation (Chapter 39) : Deliberate trading designed to move prediction market prices away from true probabilities for strategic or financial advantage. Research shows that prediction markets are relatively resilient to manipulation due to the profit incentive for corrective trading. See also: wash trading, moral hazard.

Mechanism design (Chapter 40) : A field of economics that studies how to design rules and incentive structures to achieve desired outcomes. Prediction markets and futarchy are applications of mechanism design. See also: futarchy, incentive compatibility.

Merton model (Chapter 15) : An extension of the Black-Scholes framework that accounts for continuous-time portfolio optimization and intertemporal hedging. The Merton model has analogies to dynamic trading in prediction markets. See also: Black-Scholes model.

Model drift (Chapter 26) : The degradation of a predictive model's performance over time as the statistical relationship between features and outcomes changes. Monitoring for model drift is an essential MLOps practice. See also: MLOps, regime detection.

Normal distribution (Chapter 8) : The Gaussian probability distribution, characterized by its mean and variance. Many statistical techniques used in prediction market analysis assume approximate normality. See also: central limit theorem.

Option (Chapter 15) : A financial derivative that gives the holder the right but not the obligation to buy or sell an asset at a specified price. Binary prediction market contracts share structural similarities with binary (digital) options. See also: Black-Scholes model, event contract.

Polygon (Chapter 18) : A Layer 2 scaling solution for Ethereum that provides faster and cheaper transactions. Polymarket migrated to Polygon to reduce gas costs for its users. See also: Layer 2, Ethereum, Polymarket.

Prediction market (Chapter 1) : A speculative market designed to aggregate information about the probability of future events by allowing participants to buy and sell contracts whose payoffs depend on event outcomes. See also: information aggregation, event contract, wisdom of crowds.

Probability weighting (Chapter 11) : The phenomenon, documented in prospect theory, that people systematically distort probabilities when making decisions, overweighting small probabilities and underweighting large ones. See also: prospect theory, favorite-longshot bias.

Regularized regression (Chapter 22) : Regression models that incorporate a penalty term on coefficient magnitudes (L1 for Lasso, L2 for Ridge) to prevent overfitting. See also: regularization, logistic regression.

Resolution source (Chapter 2) : The authoritative data source or entity that determines the outcome of a prediction market event. Examples include official government statistics, trusted news organizations, or decentralized oracle networks. See also: resolution, oracle.

Risk-reward ratio (Chapter 30) : The ratio of the potential loss to the potential gain on a trading position. Favorable risk-reward ratios are a hallmark of disciplined prediction market strategies. See also: edge, expected value.

Sell order (Chapter 4) : An instruction to sell a specified number of contracts at a given price. See also: limit order, market order.

Sentiment analysis (Chapter 22) : The use of natural language processing to extract subjective information (opinions, emotions, attitudes) from text data. Sentiment signals from news and social media can be features for prediction market models. See also: feature engineering, transformer.

Signal (Chapter 22) : A feature or data source that contains genuine predictive information about the outcome of a prediction market event, as opposed to noise. See also: feature engineering, edge.

Stop-loss (Chapter 32) : An order type that automatically closes a position when the price moves against the trader beyond a specified threshold, limiting potential losses. See also: risk of ruin, drawdown.

Superforecaster (Chapter 13) : A term coined by Philip Tetlock for individuals who demonstrate consistently superior probabilistic forecasting ability. Superforecasters exhibit excellent calibration, frequent belief updating, and cognitive flexibility. See also: Tetlock, calibration, Good judgment.

Tick size (Chapter 4) : The minimum price increment at which a prediction market contract can trade. For example, Kalshi contracts trade in $0.01 increments. See also: order book, CLOB.

Token (Chapter 16) : A digital asset on a blockchain, often representing a unit of value, governance right, or outcome claim. Prediction markets use various token types for trading, liquidity provision, and governance. See also: ERC-20, outcome token, governance token.

Turing completeness (Chapter 16) : The ability of a computational system to simulate any Turing machine, meaning it can execute arbitrary logic. Ethereum's smart contract language (Solidity) is Turing complete, enabling complex prediction market logic. See also: Ethereum, smart contract, Solidity.

Variance (Chapter 29) : A statistical measure of the dispersion of returns around their expected value. Lower variance is generally preferred, holding expected return constant. See also: diversification, Sharpe ratio, volatility.

Weighted average (Chapter 13) : An aggregation method that assigns different weights to individual forecasters based on factors such as past accuracy, expertise, or information content. See also: aggregation, extremizing, ensemble.

This glossary covers the principal terms from all 42 chapters of the book. Readers seeking fuller treatment of any concept should consult the referenced chapter.