Quantis Trade
← Back to all metrics

Metrics / Alpha

Alpha (annualised)

4 min read · Metric

The return your strategy delivers after stripping out everything the factor model would have given you anyway.

What it is

Alpha is the intercept of the factor regression — the part of your strategy's return that the named factors (market, size, value, momentum, quality, investment) don't explain. Annualised, it represents pure edge over and above what passive factor exposure would have produced.

Quantis reports alpha against the FF5+momentum model. CAPM alpha (vs market alone) is structurally larger because it lumps every other factor into the residual — don't compare CAPM alpha to multi-factor alpha; they're different numbers.

Formula

# OLS regression: strategy = alpha_period + sum(beta_i * factor_i) + epsilon
# Then annualise the intercept:
alpha_annualised = alpha_period * periods_per_year

# Significance test
t_stat = alpha / standard_error_of_alpha
# |t_stat| > 2 ≈ statistically significant at 5% (two-sided)

Typical ranges

  • Most active strategies: -2% to +2%, t-stat usually below 2.
  • Strategies with genuine edge: +2% to +5% with t-stat > 2.
  • Above +5% sustained: rare. Usually a regime tailwind, look-ahead bias, or hidden factor exposure the FF5+UMD model doesn't capture.
  • Negative alpha: the strategy underperforms what its factor mix would have produced. Common — most active management has negative alpha after fees.

Common mistakes

  • Not checking the t-stat. A 4% annualised alpha sounds great until you see the t-stat is 0.6 — meaning the estimate is consistent with no edge at all. The bar is |t| > 2; without that, the alpha is sample noise.
  • Comparing alphas across factor models. CAPM alpha vs FF5+UMD alpha for the same strategy will differ — usually by a lot. Always state which model.
  • Reading alpha as risk-adjusted return. It's not — it's factor-adjusted return. Two strategies with the same alpha can have very different volatility and drawdown profiles.
  • Confusing alpha with active return. Active return is strategy minus benchmark. Alpha is strategy minus what factors predict. They're not the same — a strategy with high active return can have zero alpha if all the active return came from a factor tilt.

What the platform flags

Quantis displays annualised alpha alongside its t-stat and the R² of the factor regression. A common pattern: alpha looks meaningful (3–4% annualised) but t-stat is 1.2 — the result is consistent with zero alpha given the sample size. Trust the t-stat over the headline number. Quantis colour-codes the alpha pill amber when |t-stat| < 2 to make this explicit.

Further reading