Quantis Trade
← Back to all metrics

Metrics / Beta

Beta

3 min read · Metric

How sensitive your strategy is to its benchmark. The starting point for separating market exposure from genuine alpha.

What it is

Beta is the slope coefficient from an OLS regression of strategy returns on benchmark returns. A beta of 1.0 means a 1% benchmark move is associated with a 1% strategy move on average; 0.5 means half as sensitive; 1.5 means 50% more sensitive (often via leverage or stock-picking concentration in high-beta names).

Beta isn't a verdict — high beta isn't bad and low beta isn't good. It's context. A backtest with high CAGR and beta of 1.5 might just be running a leveraged long; the “outperformance” is mechanical beta scaling, not edge.

Formula

# OLS regression: strategy_return = alpha + beta * benchmark_return + error
beta = covariance(strategy, benchmark) / variance(benchmark)

# Equivalently:
beta = correlation(strategy, benchmark) * (std(strategy) / std(benchmark))

Quantis runs the regression on the same frequency as the strategy returns (typically daily) and reports the resulting beta plus its t-stat.

Typical ranges

  • Long equity strategies vs SPY: 0.7–1.3.
  • Defensive strategies (low-vol, quality): 0.6–0.85.
  • Growth/momentum tilts: 1.0–1.4.
  • Market-neutral long-short: 0 ± 0.2.
  • Levered strategies: proportional to leverage (2x leverage → ~2.0 beta).
  • Negative beta: the strategy moves opposite to the benchmark. Tail-hedge funds and short-bias strategies; rare in long-only systematic backtests.

Common mistakes

  • Comparing betas across benchmarks. Beta vs SPY and beta vs MSCI World are different numbers. Always state the benchmark you regressed against.
  • Ignoring R². A beta of 1.5 with R² = 0.05 is statistically meaningless — the regression isn't explaining most of the variance, so the beta point estimate is noise.
  • Confusing beta with leverage. They're related but distinct. A leveraged long has high beta; a concentrated stock pick can also have high beta without leverage.

What the platform flags

Quantis shows two betas: the simple OLS-vs-benchmark slope (against SPY by default) and the FF5+momentum market beta from the multi-factor regression. They should agree closely; large divergence (e.g. 1.0 vs 0.5) means the simple beta is being inflated by exposure to other factors that the multi-factor model strips out.

Further reading