WhitmanTrading

Factor Investing: Sorting by a Number

Factor investing builds a portfolio by sorting shares on a measurable characteristic, such as cheapness, size, recent strength, profitability or low volatility, rather than by company-by-company judgement. Each characteristic is a pattern found in past data, and every one has long stretches where it does not work.

How it works

Factor investing is sorting shares by a measurable characteristic. A portfolio is built by ranking companies on one number thought to be associated with higher returns, rather than by company-by-company judgement or by market value alone.

A candlestick chart of the site's shared price history. The headline on the chart reads: Sorting shares by a measurable characteristic.
Sorting shares by a measurable characteristic. Illustrative chart - not real market data.

Value, size, momentum, quality and low volatility. Those are the characteristics most often sorted on: cheapness against fundamentals, small company size, recent relative price strength, profitability and balance-sheet quality, and low variability of price.

A gently rising stretch of the long price series with an account equity curve beneath it. The headline on the chart reads: Value, size, momentum, quality and low volatility.
Value, size, momentum, quality and low volatility. Illustrative chart - not real market data.

Each is a pattern found in historical data. Two explanations compete. Either the characteristic marks a risk investors are compensated for bearing, or it marks a behavioural error that keeps repeating, and the numbers alone cannot settle which.

A calmly advancing stretch of the long price series with a slowly rising equity curve beneath it. The headline on the chart reads: Each is a pattern found in historical data.
Each is a pattern found in historical data. Illustrative chart - not real market data.

And a published pattern gets traded against. A characteristic that predicted returns while nobody was looking at it is a different object once a great deal of money is sorted by it. Whether the premium survives that crowd is genuinely open.

A flat, quiet stretch of the long price series with a gradually rising equity curve beneath it. The headline on the chart reads: And a published pattern gets traded against.
And a published pattern gets traded against. Illustrative chart - not real market data.

Every factor has decades where it does not work. Each of them has produced stretches of underperformance long enough that almost nobody holds through one. That is the practical problem with the whole approach, not a footnote to it.

A strongly rising stretch of the long price series with an account curve breaching its limit. The headline on the chart reads: Every factor has decades where it does not work.
Every factor has decades where it does not work. Illustrative chart - not real market data.

From the pattern to the portfolio

Hundreds have been published and few survive testing. The process that produced them, searching historical data for characteristics that predicted returns, is exactly the process that produces false ones. Overfitting is the name for that failure.

A choppy, directionless stretch of the long price series. The headline on the chart reads: Hundreds have been published and few survive testing.
Hundreds have been published and few survive testing. Illustrative chart - not real market data.

And the definition changes the result completely. Cheapness can be measured against earnings, book value, cash flow or sales, and each choice produces different holdings. Two funds tracking the same named factor can own substantially different portfolios.

A declining stretch of the long price series. The headline on the chart reads: And the definition changes the result completely.
And the definition changes the result completely. Illustrative chart - not real market data.

The fund fee is subtracted from the factor. Exposure is normally bought through a fund, so the annual charge, the turnover from rebalancing and the spreads paid on it all come out of whatever premium exists before you see any of it.

A 72-bar candlestick section of the shared price history with an account curve shown with and without fees. The headline on the chart reads: The fund fee is subtracted from the factor.
The fund fee is subtracted from the factor. Illustrative chart - not real market data.

Some factors only work in names too thin to buy. An effect concentrated in small cap shares can fade when a large fund tries to trade it, because the volume is not there at the prices the backtesting assumed.

A candlestick chart with a volume histogram beneath it, with the volume histogram emphasised. The headline on the chart reads: Some factors only work in names too thin to buy.
Some factors only work in names too thin to buy. Illustrative chart - not real market data.

The evidence is measured in decades. Quantitative analysis of a factor needs a very long sample before it says anything, which means the evidence you are leaning on and the horizon you must hold for are the same length.

A long-horizon candlestick view of the same price series. The headline on the chart reads: The evidence is measured in decades.
The evidence is measured in decades. Illustrative chart - not real market data.

In practice

And the unwind can arrive in a single week. A crowded factor can reverse quickly, with an opening gap doing most of the damage before anybody rebalances. Slow to build, quick to come apart.

A candlestick series containing several opening gaps, with the largest opening gap marked. The headline on the chart reads: And the unwind can arrive in a single week.
And the unwind can arrive in a single week. Illustrative chart - not real market data.

There is no stop on a factor exposure. A stop loss makes no sense here, because the whole proposition is holding through the bad stretch. On this site’s shared 576-bar history, 95% of bars sit below a prior peak and the longest run below one is 73 bars.

A declining stretch of the long price series, with the entry price and the level at which a stop would trigger drawn as horizontal lines. The headline on the chart reads: There is no stop on a factor exposure.
There is no stop on a factor exposure. Illustrative chart - not real market data.

Every round trip costs 2% of a bar. A round trip on that same history is 0.0098 price units: 2% of a median bar’s range and 45% of the smallest bar. A factor fund pays something like that at every rebalance, not once.

A candlestick chart of the site's shared price history, annotated with the round-trip cost. The headline on the chart reads: Every round trip costs 2% of a bar.
Every round trip costs 2% of a bar. Illustrative chart - not real market data.

Compare the fee against a plain index fund first, not last. Index funds and broad exchange-traded fund holdings set the benchmark charge, so ordinary ETF investing is the right yardstick. The factor version costs more, and that difference is the price of the whole idea.

And check what you already own. A factor fund held beside a broad index fund usually has high correlation with it and carries ordinary market beta as well, so the tilt is a smaller part of the position than it looks.

Writing the exit before the entry

The decision that matters is made before you buy. How long will you hold this exposure, and what would actually count as evidence that it has stopped working? Both questions are answerable in advance and almost unanswerable during the lag.

During a lag every answer is contaminated. The underperformance itself supplies the argument for selling, and that argument always sounds like analysis rather than discomfort. Which is why the holding period has to be written down while you are calm.

So write two sentences. The first is the minimum period you will hold regardless of results. The second is the specific condition that would end it: a change in how the fund defines the factor, a rise in its charge, a merger into something else, or the exposure no longer being the one you bought.

Then leave both alone. A rule revised in the middle of a lag is not a rule. If you cannot write either sentence honestly today, that is your answer on whether to buy the fund at all.

What factor investing is not

When it fails

A sideways, range-bound candlestick series. The headline on the chart reads: In a flat decade the fee is the only certainty.
In a flat decade the fee is the only certainty. Illustrative chart - not real market data.

The exposure can be perfectly sound and still impossible to hold. These are the cases that do the damage:

The original data

A strongly rising stretch of the long price series, cut short at the decision bar. The headline on the chart reads: The factor has lagged ten years. Broken?
The factor has lagged ten years. Broken? Illustrative chart - not real market data.

The scan is in research/broker-coverage.json. Across the 31,760 trading and investing videos in research/search-study-corpus.jsonl, 3 titles mention factor investing, from 3 channels, at a median of 50,285 views and a maximum of 62,521. Two titles mention overfitting, from 2 channels, median 299 views.

Factor research is the single largest source of data-mined results in finance. The concept that decides whether a published factor was ever real is discussed in two videos out of that whole corpus. “Monte carlo”, “expectancy” and “risk of ruin” appear zero times each.

The individual factors are covered far better than the idea behind them. “Value investing” appears in 78 titles from 52 channels at a median of 13,135 views, “small cap” in 16 titles from 14 channels at 8,582, and “momentum stock” in 7 titles from 6 channels at 908.

research/series-measurements.json, built by site/measure_series.py, holds the fee arithmetic. Compounding an annual charge alone over thirty years, with no return assumption at all, 5 basis points removes 1.5% of the pot and 20 removes 5.8%. A basis point is one hundredth of a percentage point.

At 75 basis points it is 20.2%, and at 150 it is 36.5%. The fee is certain and the premium is not, which is the honest comparison to make. Before buying any factor fund, write down your holding period and your abandonment condition, then put its charge next to a plain index fund’s as the first step, not the last.

Value investing is the oldest of these characteristics and the one that tests patience hardest, which makes it the fairest test of whether you can hold a factor at all. A momentum stock screen sorts on the opposite signal, which is why a blend of the two so often ends up with conviction nowhere. Overfitting is the failure that decides whether a published factor was ever there, and it is the first question to ask about a new one.

What I actually do

I have never held a factor exposure long enough to find out whether it worked. What happens is that the lagging years arrive, the reasoning starts to feel dated, and something else is visibly doing better. By the time I sold, I could always produce a tidy argument for why that particular factor was finished. The argument was never the reason; the discomfort was.

— Michael Whitman

This page is educational, not financial advice. Test every idea on your own charts before risking money.