Alpha is the label for performance that is not easily explained by riding the market. In a simple factor story, if beta captures "how much market you owned," alpha captures "what was left over" after accounting for that exposure.
The one-line intuition
After adjusting for benchmark exposure (beta), alpha is excess return attributable to skill, selection, timing, or factors you did not model - not magic, just the residual.
Where you see it
In a linear regression of strategy returns on a benchmark:
strategy return ≈ alpha + beta × benchmark return
Alpha is the intercept (often annualized in reports). Beta is the slope.
On tear sheets, alpha may be reported versus SPY, a custom benchmark, or a blended index. The number changes with the benchmark choice.
Alpha vs "beating the market"
| Idea | Meaning |
|---|---|
| Raw outperformance | Strategy return minus benchmark return over a period |
| Risk-adjusted alpha | Outperformance after scaling for beta and sometimes other factors |
A strategy can beat the index in a bull market with beta > 1 while contributing little true alpha. Always read beta alongside headline returns.
Why backtest alpha disappoints live
Common reasons alpha shrinks:
- Overfitting - rules tuned to noise in one historical window (backtest red flags).
- Costs - commissions and slippage were understated.
- Capacity and liquidity - backtests assume fills you cannot get at size.
- Regime change - relationships that held for years stop working.
- Survivorship and data quirks - especially in narrow universes.
Treat strong in-sample alpha as a hypothesis to stress-test, not a promise.
Alpha and systematic rules
Rules-based strategies can still seek alpha through:
- Selection (which names enter the portfolio)
- Timing (when to be invested)
- Weighting (how capital splits across signals)
The question is whether that edge survives realistic execution and future data.
Limits
- Alpha is model-dependent (benchmark, frequency, regression window).
- Single-factor alpha ignores value, size, momentum, and other exposures.
- Short track records produce meaningless precision.
Alpha is a useful summary statistic when defined consistently. Combine it with drawdown, volatility, and a clear story about what your rules actually do.
