Strategies · Library · ETFs
§4.3Factor
R-squared
Rank ETFs by how much of their return the factor regression explains (low fit means more independent behaviour) and by alpha, then favour low-fit, high-alpha ETFs. Low fit is read as a sign of active selection.
- Asset class
- ETFs
- Held for
- Months
- Data
- Needs data ARCANE lacks
- Symphony
- Cannot be built as a symphony
The rule
- Signal
- R-squared and alpha of a factor regression over about a year
- Entry and exit
- Buy the group with low R-squared and high alpha; sell the opposite group; rebalance periodically.
- Sizing
- Equal within each group
How it fails
- Needs a regression engine
- Short leg
- Weak evidence for ETFs
Data it needs
- inputs ARCANE does not hold
Needs factor return series and regressions, which ARCANE does not have.
How ARCANE uses it
Not drafted directly; a note on low-correlation ETFs could inform a diversification comment.
Needs regression statistics and a short leg, which the language lacks.
Filed under the factor card
Own companies that are small, cheap, profitable and invest conservatively, against their opposites, and rebalance once a year.
A paper plan from this card holds 63 to 365 days, stops at 2× the expected move and uses at most 75% of your risk budget. The card
Source: Kakushadze and Serur (2018), 151 Trading Strategies, §4.3, p. 64 of the SSRN version. Described in our words. Factor is ARCANE’s filing, not the book’s.