Factor investing
Factors are recurring patterns that explain return differences between stocks, such as company size or valuation level. A factor regression decomposes a portfolio return into these building blocks and shows how much remains as its own contribution.
- Barely explained
- The portfolio movement has little to do with the tested factors. Common for highly concentrated holdings or those with large allocations outside the equity market. Betas and alpha should be read with caution here.
- Partly explained
- A noticeable own share remains. Typical for portfolios holding a few single stocks alongside a broad core.
- Well explained
- The usual range for mixed portfolios. Most of the movement comes from the market and known patterns.
- Almost fully explained
- A typical market blend. Practically all movement traces back to factors, which is the expected finding for broad ETF portfolios.
How it is calculated
R[portfolio] − R[rf] = α + β₁·market + β₂·size + β₃·value + … + ε
Every factor is represented by an investable ETF rather than a theoretical data series. Gold, crypto and bonds are included because otherwise their movement would wrongly appear as alpha.
- Market
- MSCI World
- Style
- Size, value
- Style
- Momentum, quality
- Other
- Min vol, gold, crypto, bonds
What the number does not tell you
- Below 36 common months the result is rough orientation. The regression needs data points. The fewer common months portfolio and factors share, the more the estimated betas move around. Below three years, even the signs are often chance.
- Factors overlap. Value and size have historically run partly in parallel. From a VIF of 5, their betas can no longer be cleanly separated: the model knows a tilt is there but not which factor owns it.
- A premium is not a promise. Factor premiums show what a pattern delivered over the measured period. Value lagged for more than a decade before it returned. Premiums vanish for years and reappear.
- The proxy is not the factor. A momentum ETF reproduces momentum with lag, costs and trading rules. The measured beta describes proximity to that ETF, not to the academic factor.
Related metrics
How Evergrova calculates it
The factor page shows exposure per factor with confidence interval and t-statistic, the decomposition of excess return, rolling betas over time, plus model quality, residual volatility and degrees of freedom. Proxy ETFs can be adjusted per portfolio.
Calculation: monthly multifactor regression of portfolio excess returns on investable factor ETFs; market proxy MSCI World.
View the demo portfolioCommon questions
What is a good R-squared in a factor regression?
High does not mean good, it means typical. An R-squared of 0.90 says the portfolio is explained almost entirely by the market and known patterns. For broad ETF portfolios that is the expected finding, not a shortcoming.
Why are gold and crypto included as factors?
Because otherwise their movement lands in alpha. A portfolio with 10 % gold during a gold rally would look like skill without that factor. With it, the contribution is visibly attributable to the asset class.
Is factor investing worth it?
That question cannot be answered here, because it depends on horizon and staying power. What the regression does show is something else: which tilts your portfolio already carries, often without having chosen them deliberately.
Last reviewed: 2026-08-08