Terminology and Definitions

Cross-Asset Alpha Engine Glossary

Alpha Generation Terms

Alpha: Excess return generated by a trading strategy relative to a benchmark, representing the value added by active management.
Cross-Asset Alpha: Alpha generated by exploiting relationships and inefficiencies across multiple asset classes (equities, bonds, commodities, currencies).
Regime-Aware Alpha: Alpha generation that adapts to different market conditions or regimes, using different models or parameters for different market environments.
Information Ratio: Risk-adjusted measure of alpha generation, calculated as excess return divided by tracking error (standard deviation of excess returns).

Market Regime Terms

Market Regime: Distinct periods in financial markets characterized by different risk-return dynamics, volatility patterns, and asset correlations.
Hidden Markov Model (HMM): Statistical model assuming that market observations are generated by an underlying, unobservable regime state that follows a Markov process.
Regime Detection: Process of identifying and classifying different market regimes using statistical or machine learning methods.
Regime Transition: The process of moving from one market regime to another, often triggered by economic events or structural changes.
State Space Model: Mathematical framework where the system state (regime) is not directly observable but can be inferred from observable variables.

Feature Engineering Terms

Technical Features: Quantitative indicators derived from price and volume data, including momentum, volatility, and mean reversion signals.
Daily Microstructure-Inspired Features: Features inspired by microstructure concepts but computed from daily OHLCV bars only. Includes VWAP deviations, volume anomalies, and daily price patterns. Note: True intraday trading patterns and bid-ask dynamics require intraday/tick data, which is not used in the current experiment.
Cross-Asset Features: Indicators that capture relationships between different asset classes, such as correlations, volatility spillovers, and risk sentiment.
Feature Importance: Measure of how much each feature contributes to model predictions, typically calculated using methods like Gini importance or permutation importance.
Z-Score Normalization: Statistical technique to standardize features by subtracting the mean and dividing by standard deviation, ensuring features have zero mean and unit variance.

Statistical and ML Terms

Random Forest: Ensemble machine learning method that combines multiple decision trees to make predictions, providing robustness and feature importance rankings.
Ensemble Method: Machine learning technique that combines predictions from multiple models to improve overall performance and reduce overfitting.
Walk-Forward Validation: Time series validation technique where models are trained on historical data and tested on subsequent out-of-sample periods.
Cross-Validation: Model validation technique that divides data into multiple folds to assess model performance and prevent overfitting.
Overfitting: Phenomenon where a model performs well on training data but poorly on new, unseen data due to excessive complexity.

Risk Management Terms

Value at Risk (VaR): Statistical measure estimating the maximum potential loss of a portfolio over a specific time horizon at a given confidence level.
Expected Shortfall (ES): Risk measure that estimates the expected loss beyond the VaR threshold, also known as Conditional Value at Risk (CVaR).
Maximum Drawdown: Largest peak-to-trough decline in portfolio value, representing the worst-case loss scenario during the analysis period.
Sharpe Ratio: Risk-adjusted return measure calculated as excess return divided by standard deviation of returns.
Market Neutrality: Portfolio construction approach that maintains approximately zero net market exposure (beta ≈ 0) to isolate alpha from market movements.

Portfolio Construction Terms

Position Sizing: Process of determining the appropriate allocation to each asset in a portfolio based on expected returns, risk, and constraints.
Risk Parity: Portfolio construction approach where each asset contributes equally to total portfolio risk, typically achieved by inverse volatility weighting.
Kelly Criterion: Mathematical formula for optimal position sizing that maximizes long-term growth rate based on win probability and payoff ratios.
Gross Exposure: Sum of absolute values of all portfolio positions, representing total capital deployed regardless of direction.
Net Exposure: Sum of all portfolio positions considering direction (long minus short), representing overall market exposure or beta.

Market Data Terms

OHLCV: Standard market data format containing Open, High, Low, Close prices and Volume for each time period.
VWAP (Volume Weighted Average Price): Average price weighted by volume, representing the average execution price for the trading period.
Bid-Ask Spread: Difference between the highest price a buyer is willing to pay (bid) and the lowest price a seller is willing to accept (ask).
Market Impact: Price movement caused by executing a trade, typically modeled as a function of trade size and market liquidity.
Slippage: Difference between expected execution price and actual execution price, caused by market movement and liquidity constraints.

Asset Class Definitions

SPY: SPDR S&P 500 ETF Trust, tracking the S&P 500 index and representing large-cap US equity exposure.
QQQ: Invesco QQQ Trust, tracking the NASDAQ-100 index and representing technology-heavy large-cap growth stocks.
IWM: iShares Russell 2000 ETF, tracking small-cap US equity exposure.
VIX: CBOE Volatility Index, measuring implied volatility of S&P 500 options and serving as a "fear gauge" for market sentiment.
TLT: iShares 20+ Year Treasury Bond ETF, representing long-term US government bond exposure and interest rate sensitivity.
GLD: SPDR Gold Trust, providing exposure to gold prices and serving as an inflation hedge and safe-haven asset.
DXY: US Dollar Index, measuring the value of the US dollar against a basket of major foreign currencies.
USO: United States Oil Fund, tracking crude oil prices and representing commodity exposure.

Performance Metrics

Annualized Return: Return scaled to represent performance over a full year, calculated as (1 + period_return)^(252/periods) - 1 for daily data.
Volatility: Standard deviation of returns, typically annualized by multiplying daily volatility by √252.
Win Rate: Percentage of trading periods with positive returns.
Calmar Ratio: Risk-adjusted return measure calculated as annualized return divided by maximum drawdown.
Sortino Ratio: Modified Sharpe ratio that only considers downside volatility in the denominator, focusing on harmful volatility.

Execution Terms

TWAP (Time Weighted Average Price): Execution strategy that spreads trades evenly over time to minimize market impact.
Implementation Shortfall: Difference between the decision price and the final execution price, including market impact and timing costs.
Participation Rate: Percentage of total market volume that a trading algorithm is allowed to consume during execution.
Transaction Costs: Total cost of executing trades, including commissions, bid-ask spreads, market impact, and opportunity costs.

Statistical Terms

Autocorrelation: Correlation of a time series with a delayed copy of itself, measuring the persistence of trends or mean reversion.
Stationarity: Statistical property where the mean, variance, and autocorrelation structure remain constant over time.
Heteroskedasticity: Condition where the variance of errors is not constant across observations, common in financial time series.
Multicollinearity: High correlation between predictor variables that can cause instability in model coefficients.
P-Value: Probability of observing results at least as extreme as those observed, assuming the null hypothesis is true.

Backtesting Terms

In-Sample Period: Historical data used to train and optimize models, also known as the training set.
Out-of-Sample Period: Historical data reserved for testing model performance, simulating real-world deployment conditions.
Look-Ahead Bias: Error in backtesting where future information is inadvertently used to make historical decisions.
Survivorship Bias: Bias that occurs when analysis only includes assets that survived the entire period, ignoring delisted or failed assets.
Data Snooping: Bias that results from testing multiple strategies on the same dataset and selecting the best-performing one.

This comprehensive glossary provides definitions for all key terms used throughout the Cross-Asset Alpha Engine documentation and analysis.