Skip to content
All simulations

Beta and the market

Regress a real stock against the S&P 500 to split its returns into market, alpha and noise.

Window, ending Sep 2026
4.00%

What a short-term Treasury pays. Returns are measured above it.

5.50%

Extra yearly return investors expect for holding the whole market.

Beta2.21
Alpha, a year+34.7%Past, not promised
R², explained by market48%
Correlation0.69

60 months from Oct 2021. Each dot is one month. The slope of the line is beta.

Over these 60 months, when the S&P 500 moved 1%, Nvidia moved about 2.21% on average. The market explains 48% of its ups and downs; the rest is news specific to the company. CAPM says an asset with this beta should earn about 16.2% a year.

Where the risk comes from

Market risk 48% Specific risk 52%
Total volatility50%a year
Specific volatility36%diversifiable

Specific risk can be diversified away by holding many assets, so CAPM says it earns no reward. Only market risk, measured by beta, is paid for.

Beta is not a constant

Beta measured over each trailing three years of Nvidia.

The dashed line is a beta of 1, moving exactly with the market.

Try this

  • Compare Coca-Cola with Nvidia. Which one is the defensive stock, and how can you tell from the dots?
  • Pick long-term Treasuries or gold. What does a beta near zero, or below it, mean for a portfolio?
  • Watch Tesla's rolling beta. Would a single beta number have described it well?

The math

Ri−rf=α+β (Rm−rf)+ε,β=Cov⁡(Ri,Rm)Var⁡(Rm)R_i - r_f = \alpha + \beta\,(R_m - r_f) + \varepsilon, \qquad \beta = \frac{\operatorname{Cov}(R_i, R_m)}{\operatorname{Var}(R_m)}
RiR_i
return of the asset
RmR_m
return of the market (S&P 500)
rfr_f
risk-free rate
β\beta
sensitivity to the market
α\alpha
return the market doesn't explain
ε\varepsilon
specific, diversifiable noise

Monthly adjusted prices, dividends reinvested. Historical results describe the past and are not a forecast or a recommendation.