Definition

Alpha is the component of a portfolio's return that a specified risk model does not explain. Its meaning depends on the benchmark, risk factors, return frequency and estimation window. In casual usage, “alpha” may mean any return above a benchmark, but that shorthand is different from regression alpha.

Alpha and beta formula

A single-factor excess-return regression is:

portfolio return - risk-free return = alpha + beta × (market return - risk-free return) + residual

Alpha is the estimated intercept; beta measures sensitivity to the selected market factor. The intercept uses the same periodic units as the returns. Annualization and statistical uncertainty need to be stated separately. A multi-factor model can explain returns that appear as alpha under a simpler one-factor model.

A simple active return is portfolio return - benchmark return each period. Subtracting two CAGRs does not estimate the intercept above. It also does not control for leverage, defensive holdings, sector concentration or time-varying exposure.

A U.S. example of the distinction

The U.S. strategy comparison reports momentum-system CAGR of 29.9% against QQQ at 20.7% over May 2016 to October 2026. That is an annualized return comparison. It is not a measured regression alpha.

The system holds leveraged ETFs in a portfolio that also uses defensive assets and a QQQ regime gate. Its raw return difference can reflect exposure and allocation decisions as well as its signal. To estimate alpha, use aligned periodic excess returns and explicitly choose the explanatory model. A different model can produce a different intercept.

How to evaluate an alpha claim

Ask which factors are included, how cash and costs are treated, and whether exposures are stable. Inspect confidence intervals and the sample size. Check how many strategies or factor combinations were tried before the reported result was selected. An attractive intercept from a short or selected sample can be unreliable.

The factor research overview separates common signal families. For reproducible factor definitions, Fama and French's data library documents research portfolio construction. Those definitions provide model context; they do not establish alpha for a particular strategy.

Common pitfalls

Calling every positive benchmark difference “alpha” hides risk-model choices. A higher-risk strategy can outperform in a rising market without a positive risk-adjusted intercept. An alpha estimate also depends on its window and may change when another explanatory factor is added.

Frequently asked questions

Is positive alpha proof of skill?

No. It is an estimate conditional on a model and sample. Omitted exposures, data errors, costs and strategy selection can affect it.

How are alpha and beta different?

Beta is a sensitivity coefficient in the model. Alpha is its intercept. They describe different parts of the same fitted relationship and neither is guaranteed to remain stable.

See information ratio, CAGR, Sharpe ratio, backtesting and the quant glossary.

For education and research only. This is not investment advice. Historical performance does not guarantee future results.

Last updated: 2026-10. Author: FinLab Quant Research.