Alpha

Alpha is the intercept term in a linear regression of a portfolio's excess returns against a benchmark's excess returns. It represents the return attributable to manager skill rather than market exposure.

• Jensen's Alpha: The most common measure, derived from the Capital Asset Pricing Model (CAPM).
• Formula: α = Rp - [Rf + β × (Rm - Rf)], where Rp is portfolio return, Rf is the risk-free rate, and β is market sensitivity.
• A positive and statistically significant alpha indicates the manager generated returns beyond what their beta exposure would predict.

A critical distinction exists between true alpha and compensated exposure to known risk premia. Many strategies marketed as alpha are actually harvesting alternative beta.

• Pure Alpha: Idiosyncratic return uncorrelated with any known systematic factor. It is zero-sum and scarce.
• Alternative Beta: Return from systematic factors like Value, Momentum, or Carry. These are well-documented risk premia, not skill.
• Orthogonalization is the process of hedging out factor exposures to isolate the residual, pure alpha component of a signal.

Active management is a zero-sum game before costs. For every manager generating positive alpha, another must be generating negative alpha of equal magnitude.

• The aggregate portfolio of all active investors is the market portfolio itself.
• Gross alpha sums to zero across all active participants.
• After transaction costs, fees, and taxes, the net alpha across all active managers is negative.
• This mathematical constraint makes persistent positive alpha exceptionally rare and valuable.

The Information Ratio (IR) measures the consistency of alpha generation relative to the risk taken to achieve it. It is the primary metric for evaluating active management skill.

• Formula: IR = (Portfolio Return - Benchmark Return) / Tracking Error.
• An IR above 0.5 is generally considered good; above 1.0 is exceptional.
• The IR is directly related to the Information Coefficient (IC) and the breadth of independent bets: IR ≈ IC × √Breadth.
• A high Sharpe Ratio with a low IR suggests returns are driven by market beta, not alpha.

Alpha signals are not static. They exhibit a half-life as they are discovered and arbitraged away by competing investors.

• Alpha Decay Profile: The rate at which a signal's predictive power diminishes post-discovery.
• Capacity: The maximum dollar amount that can be deployed before the strategy's own trading moves prices and erodes its alpha.
• High-frequency, high-Sharpe strategies typically have low capacity and rapid decay.
• Slower, structural risk premia have higher capacity but lower Sharpe ratios.

With thousands of strategies being tested, many will show impressive backtested alpha purely by random chance. Rigorous statistical methods are required to separate skill from luck.

• Multiple Testing Correction: The False Discovery Rate (FDR) framework controls for the expected proportion of false positives among discovered signals.
• Deflated Sharpe Ratio: Adjusts the Sharpe Ratio for the expected maximum performance that would arise from data snooping.
• Walk-Forward Analysis: Validates alpha stability by testing on truly unseen, sequential out-of-sample periods rather than a single holdout set.

The primary measure of a factor's predictive skill, calculated as the correlation between a signal's forecasted values and subsequent realized returns. A higher IC indicates greater forecasting accuracy.

• Spearman Rank IC: Measures monotonic relationship, robust to outliers
• Pearson IC: Measures linear correlation, sensitive to magnitude
• A monthly IC above 0.05 is generally considered strong in equity markets

A measure of risk-adjusted active return, calculated as the ratio of a portfolio's excess returns over a benchmark to the standard deviation of those excess returns. The IR reveals whether a manager's alpha is consistent or merely lucky.

• IR = Alpha / Tracking Error
• An IR above 0.5 is considered good; above 1.0 is exceptional
• Directly linked to the breadth of independent bets placed

A portfolio construction technique that hedges out market exposure by ensuring the weighted average beta of long and short positions equals zero. This isolates pure alpha from broad market movements.

• Eliminates the equity risk premium from active returns
• Critical for market-neutral statistical arbitrage strategies
• Can be achieved via dollar-neutral or beta-neutral weighting schemes

A mathematical process of transforming a target factor signal to be uncorrelated with known risk premia. This ensures the resulting alpha is not a repackaging of value, momentum, or other established factors.

• Uses linear regression to strip out common factor exposures
• Prevents paying active fees for passive factor beta
• Essential for demonstrating genuine skill to allocators

A statistical test that adjusts a strategy's Sharpe Ratio for the expected maximum performance arising purely by chance from multiple testing. It explicitly penalizes data snooping and p-hacking.

• Accounts for the number of trials attempted
• A deflated Sharpe above 1.0 provides strong statistical confidence
• Developed by Marcos López de Prado to combat backtest overfitting

A database that stores historical information exactly as it was reported on a specific past date, without subsequent revisions. This is essential for eliminating look-ahead and restatement biases in quantitative research.

• Prevents using revised earnings that weren't yet announced
• Critical for accurate backtesting of fundamental signals
• Contrasts with 'as-of-today' databases that overwrite history