A note before I start, the same one I put on everything I write here. This is a description of how factors behave and how I think about building with them. It is not advice, it is not a recommendation, and nothing in it is a view on any individual security. The table above is a simulated research illustration and carries a note on its own face saying so. The argument underneath it carries no figures of its own, on purpose, because a number without its full context is worse than no number at all.
I went back on The Money Runner with David Nelson. The conversation started where these conversations usually start, with what the market had just done, and ended somewhere more useful.
A quiet index sitting on top of a violent cross-section
The framing David opened with was the right one. The index was calm. Volatility at the index level was unremarkable. Underneath it, the dispersion between individual names was not unremarkable at all. The safest stocks and the riskiest stocks were separating by margins that do not show up when you only look at the aggregate.
This is a recurring feature of markets rather than a novelty, and it is worth saying plainly because it is easy to be lulled by it. An index is a weighted average, and averages are very good at hiding the thing you actually own. If your portfolio is a small selection out of a large universe, the aggregate tells you about the market risk you share with everybody and very little about the part that is yours alone. It tells you about the risk the average investor is carrying, which is not the same person.
What a momentum factor is actually holding
Here is the part of the conversation I care most about, because it is a construction question rather than a market call.
A momentum factor, built the ordinary way, ranks every company against the whole universe and buys the top of that ranking. That sounds neutral. It is not. Whatever theme has been running hardest will dominate the top of a universe-wide ranking, so the factor quietly loads up on it. The name says momentum. The holdings say whichever sector has been working.
So a momentum book is carrying two bets at once, and only one of them was chosen deliberately. The first is the intended bet: that recent relative strength persists. The second is an unintended sector concentration that arrives as a side effect of how the ranking was defined. When people say a factor stopped working, quite often what happened is that the second bet was the one that mattered and nobody had written it down.
That is what the quiet index was hiding. Not a calm market, but a cross-section rearranging itself underneath a single number, and a factor built on top of it inheriting whichever part of that rearrangement had been winning.
Neutralize by industry and you are looking at a different factor
The test for this is simple, and it is one of the more useful things anyone can run on their own book. Instead of ranking against the whole universe, rank within each industry and take the strongest names inside each one. Same idea, same signal, one constraint added.
Then hold the two side by side and look at what changed: the sector weights, the names that are in one list and not the other, and the shape of the two paths.
What comes out is a different portfolio, and it does not behave like the one you started with. That difference is not a flaw in either version. It gives you a sense of how large the sector bet had been, and only a sense, because ranking inside industries also changes which names you hold and how many, so the sector exposure is not the only thing that moved.
You have not improved the factor by neutralizing it. You have pulled one of the two bets away from the other, and you can now decide how much of it you want back. The separation is partial. A theme that runs across several industries is reduced this way rather than removed, because those names still sit at the top of each of the industries it spans.
That is the whole point. A concentration you did not choose is still a concentration. It will be there in the good months, where it will look like skill, and it will be there in the bad ones, where it will look like bad luck. Neither reading is right.
Two portfolios can arrive at the same place and not be the same result
David and I spent a while on the difference between a return and a risk-adjusted return, and it is worth restating without any numbers attached.
Two portfolios can finish a period at the same level. One got there in a fairly straight line. The other got there through a sequence of sharp advances and sharp reversals. On a chart that ends at the same point, these look like the same outcome. They are not the same outcome, and treating them as equivalent is one of the most expensive mistakes available to a systematic investor.
Two reasons. The first is statistical: a volatile path is evidence that the process behind it has a wider distribution of outcomes than the smooth one, which makes the next period harder to read off the last. The second is human, and it is the one that actually decides things.
The drawdown is a constraint, not a statistic
An investor who cannot hold a position does not receive its return. That sentence sounds obvious written down and is ignored constantly.
A deep decline is not merely an unpleasant number in a report. It is the event that determines whether the person still owns the strategy afterward. If the decline is severe enough that they sell, then whatever the model would have done next is irrelevant to them, because they are not in it. The return that matters is the one the investor actually collected, and that number is bounded by what they were able to sit through.
This is why I keep coming back to the same framing, which I have written about before: you win more by losing less. Not because avoiding losses is clever, but because a shallower path is one a person can stay in, and staying in is the precondition for everything else.
It also changes with who is asking. Somebody three decades from needing the money and somebody two years from it are not facing the same problem, even if they are looking at the same strategy. The strategy did not change. The constraint did.
Blending a defensive sleeve is a tradeoff, and it is not a hedge
One thing that came up is the idea of holding something defensive alongside something aggressive, rather than moving entirely to cash or to the benchmark when a position starts to feel uncomfortable.
The honest version of that idea has two halves and most people only repeat the first. The first half: mixing in a lower volatility sleeve does reduce the depth of the declines. The second half: it also reduces the upside, and the reduction is real, not a rounding error. There is no configuration in which you keep the full advance and remove the drawdown. If someone shows you one, look harder at how it was tested.
There is a second caveat I want to be explicit about, because it gets glossed over. A low volatility allocation is not a hedge. It behaves defensively relative to the rest of the equity market, which is not at all the same as being protected from it. In a genuine market-wide dislocation, defensive equities fall too. What a defensive sleeve does is dampen the sector and style swings of the book it sits beside. It does that by carrying sector and style tilts of its own, which is the construction problem from earlier in this piece arriving from the other direction. Worth knowing before you decide the sleeve is the neutral part of the portfolio. What it does not do is remove market risk, because it is still equity.
The turkey
The old turkey analogy is the cleanest way I know to say the last part.
Every day the turkey is fed. Every day the evidence supporting the theory that it will be fed tomorrow gets a little stronger. The confidence in the model peaks the day before Thanksgiving, which is precisely the moment it is most wrong.
An equity curve that has been rising is evidence about the days that already happened. It carries some information about the days ahead, which is the premise the first half of this piece rests on, and a great deal less than the length of the run makes it feel like it carries. This is not an argument that any particular run is about to end. It is an argument that the length of the run is not the thing that tells you.
What I would not say
David asked me more than once, in different ways, whether this was the turn. Whether momentum is finished, whether a different factor takes over, whether the thing everybody is watching is a bubble about to burst.
I did not answer, and I want to be clear that this was not evasiveness for its own sake. A few weeks of one factor behaving differently is not a regime change. It might be the start of one. It might equally be noise that gets forgotten by the next quarter. The people who can tell you which one it is, with confidence, in real time, are not people I can help you find.
What I do think is defensible is narrower and more boring. You can know what your book is actually holding rather than what its label claims. You can separate the bets you chose from the ones that arrived by construction. You can decide in advance how much of a decline you are able to sit through, and build for that number rather than for the return you would like to see. None of that requires predicting anything.
That is the part of this job that is available to everybody, and it is almost all of the part that matters.