Clover Capital
All news
Research8 min read

Factors are a language; honest backtests are its grammar

Value, momentum, quality and low-volatility are useful lenses, but any claim that a factor "works" is a claim about the past, and most backtests quietly assume you knew things before you did. We describe how evidence-based premia are combined with disciplined risk control, and how point-in-time data keeps the evidence honest.


The academic literature has documented a handful of robust return premia — value, momentum, quality, profitability, low volatility — that persist across markets and decades. We take that evidence seriously. We also take seriously that any single factor can underperform for years.

The systematic framework treats factors as a shared language for describing why a portfolio is positioned the way it is, not as articles of faith. Signals are combined, scored and aggregated, then passed through portfolio optimisation and explicit risk constraints before a single position is sized.

But "evidence-based" is itself a claim about the past: had you followed these rules, here is what would have happened. The integrity of that claim depends entirely on whether the data used to form it was actually knowable at each point in the simulated history. A factor is only as trustworthy as the backtest that supports it.

Two biases corrupt most research datasets. Look-ahead bias creeps in when a backtest uses information — a restated earnings figure, a delisting flag, an index reconstitution — before the market actually had it. Survivorship bias appears when the universe only contains the companies that still exist today, silently deleting the failures from history. Both make a factor look better than it ever was.

The research platform's data is bitemporal: every fact carries both the date it refers to and the date it was learned. A point-in-time query reconstructs exactly what was knowable on any past date, and index membership is stored as a slowly-changing history rather than a snapshot. The result is a backtest that cannot accidentally cheat, and a set of factor premia measured as they would genuinely have been earned.

The discipline matters most when a factor is out of favour. A systematic process lets an investor hold an evidence-based position through drawdowns that would shake a purely discretionary view — provided the original evidence still holds when it is re-examined. That re-examination is only meaningful if the evidence was honest to begin with. This is unglamorous infrastructure, but it is the difference between a strategy that works on paper and one that survives contact with live markets. We would rather discover a disappointing-but-real edge than a spectacular-but-fictional one.

The views expressed are for information only and are not investment advice.


Want to explore investment opportunities with us?

We are happy to talk through our approach with eligible investors.

Get in touch