What is a factor return, exactly?
"Value returned 4% a year" is a sentence people say without knowing what the 4% is the return of. It is the return of a specific portfolio, built by a specific recipe, and this lesson is the recipe — because a factor fund captures part of it and a screen captures a different part.
Sort, split, subtract
Take every eligible share on a date. Rank them on the characteristic — book-to-market, past twelve-month return, market capitalisation. Split the ranking into groups: tenths (deciles), fifths, or the academic convention of three groups at the 30th and 70th percentiles. Hold each group as a portfolio, usually weighted by capitalisation, for a fixed period — a month for momentum, a year for value and size in the original papers. Then re-sort and repeat.
The factor return for a period is the top group's return minus the bottom group's. Long the cheap tenth, short the expensive tenth: the spread is what "value returned" means. It is a long-short return, it has a beta near zero by construction, and it can be positive in a year the market falls and negative in a year it soars.
Three things the recipe decides
The sort variable. Book-to-market and earnings yield both measure value and disagree on a third of the names. Twelve-month return and six-month return both measure momentum. The published factor is one choice; a fund's index is another; the two can diverge for years.
The breakpoints and the weighting. Deciles concentrate the effect and the noise; thirds dilute both. Capitalisation weighting keeps the portfolio holdable and lets a few giants dominate a group; equal weighting gives the factor its full strength in the small names nobody can trade. Much of the size premium, measured equal-weighted, is a premium on micro-caps.
The rebalancing period. Value sorted once a year turns over little. Momentum sorted every month turns over most of the portfolio every year, and the trading is a cost that the published spread does not subtract.
Why a fund is not the spread
A long-only fund holds the top group and nothing else. It earns the market return plus the top half of the spread — the part that comes from cheap shares beating the market — and forgoes the bottom half, where expensive shares lag it. For most factors the two halves are not equal, and a good deal of momentum's historical spread came from the short side. A fund with value in its name and a factor with value in its name are cousins, not twins, and the long-only cousin is the only one most people can hold.
In the data
A sort needs a universe, a characteristic and prices. What you can actually hold of one is the long-only top group, and the Russell 1000 value fund is that cousin. Its whole life, from the first day its price history holds, 26 May 2000, four days after launch:
The fund is cut from the same parent index as its growth twin: on 2026-09-04 the value fund held 864 names and the growth fund 366, roughly a 70/30 split of a thousand names by value score, with some names allowed in both. That overlap is one of the recipe's choices, and the published factor makes a different one.
Try it now
- Write the recipe for a momentum factor as five numbered steps, naming the sort variable, the breakpoints, the weighting, the holding period and the subtraction. If any step is missing, the return is not defined.
- The long-only cousin against the market, at the range where its whole history is visible:
Measure 26 May 2000 to 3 September 2026 on the market. Compute the value fund's return over the same window from its two adjusted closes in the table above, and write down whether the long-only cousin beat the market over its full life.