Alpha

Alpha is what is left of your return after accounting for the return you would have got simply by being exposed to the benchmark.

If your portfolio has a beta of 1.2 and the benchmark rose, some of your gain was just that exposure showing up. Alpha strips that part out and reports the remainder, annualised.

How it is calculated

This is Jensen's alpha, from the capital asset pricing model:

alpha = (mean portfolio return − risk-free − beta × (mean benchmark return − risk-free)) × 252

Read the middle term as the return your beta entitled you to. Whatever your portfolio did beyond that is alpha. The daily result is annualised by multiplying by 252, the number of trading days in a year.

Positive alpha means the portfolio did better than its benchmark exposure alone would predict. Negative means worse.

Two assumptions worth knowing

The risk-free rate is fixed at 3%. It is a configuration constant, not a live market rate, and it appears twice in the formula. See Sharpe ratio for more on why this is a stated assumption rather than a measurement.

CAPM is a model, not a law. Alpha is defined relative to it, so it inherits every one of its simplifications. The model assumes a single source of risk that beta captures fully, which is not how portfolios actually work. In practice, a chunk of what shows up as alpha is exposure to something the benchmark does not contain: a sector, a size, a currency, a factor.

How to read it

Alpha is the measure most often over-read, so it is worth being blunt: over short windows it is mostly noise.

It is estimated from a beta which is itself estimated, across a limited number of days, against a benchmark you chose. Change the benchmark and the alpha changes. Change the window and it changes again. A positive alpha over three months is not evidence of skill, and Gylder does not present it as such.

For a benchmark-relative measure that includes how consistently you beat the benchmark rather than only by how much, see the information ratio.