This is not part of the weekly series. It is a question for readers, with a poll.

Why it exists

Everything in this newsletter comes from ten research papers about numbers that steer by the past. A paper can be checked in a few ways. Other researchers can review it. Its predictions can be graded — the flu alarm and the recession board are doing that now. And it can be put to work, in a setting where being wrong costs money.

This strategy is the third kind of check. I built it to find out whether the research holds up when real money rides on it. It is not inspired by the papers. It is made of them. The measurement lag, the blind period, the fast-against-slow gap, the loops: those are the moving parts, and each one does a specific job inside the rules. Take the research away, and there is no strategy left.

That cuts both ways, and it is meant to. If the strategy performs roughly as the research says it should, the theory behind it stands on firmer ground than it does today. Not the footing of a peer-reviewed result — that is a different test, and it is under way separately. But a working strategy on real money, doing what the papers predicted, is more than an independent researcher with AI tools can claim on his own. And if it fails, that says something about the papers too. I would rather find that out in public.

What it is

It holds four ordinary ETFs: US stocks, tech stocks, gold, and Treasury bills. It starts with equal weights, and then, every trading day, a set of rules reads the market and decides how much of each to hold. When the rules see danger building, exposure comes down. When they don’t, it goes back up. That is the whole job. There is no forecasting, no stock picking, and nothing exotic in it.

What it is not

I did not set out to build this. I went looking for something bigger and faster, and the research kept leading the other way. Every attempt to make it more aggressive — leverage, options overlays, a second sleeve, a growth dial — was tested and failed to earn its cost. The version that survived is the plain one.

So here is the honest pitch, and it is not much of a pitch. In a backtest from 2002, the strategy returned about 7.4% a year net of reasonable costs, with a worst peak-to-trough loss of about 8.5%. That is a backtest: hypothetical, with all the usual caveats. It lags the stock market in good years, by design. It would not have made anyone rich.

Its job is different. It sits between cash and the stock market: more return than Treasury bills, far less pain than owning the index outright. It is meant for the part of a portfolio you cannot afford to see cut in half. And it is meant to get through bad years without a loss that takes a decade to recover from.

There are two ways to lose money, and most people only fear one of them. The first is the crash: fast, visible, on the news. The second is slow and silent. Money left in cash to be safe loses a little of its buying power every year. After a decade, the loss is real, even though the balance never went down. This strategy is meant to sit between those two fears. It aims to stay clear of the crash, and over time to stay ahead of inflation rather than fall behind it. Not by a lot. But ahead.

For readers who judge by the ratios, here are the approximate figures from that backtest, after modeled trading costs. Volatility around 6.5%. A Sharpe ratio around 1.1, and a Sortino ratio around 1.5. A Calmar ratio around 0.85 over the whole period, measured against its single worst drawdown, and about 0.9 on the rolling three-year basis most people use.

And one figure that matters more to me than any ratio. Across twenty-four years of backtesting, measured over every possible three-year window, the strategy has never had a losing three-year period. Not one.

None of that is spectacular. It is meant to be realistic and achievable, which is a different goal. The point of the profile is the balance: the returns are moderate, and the volatility and drawdowns are small relative to them. That balance is what I set out to test, and it is what I would report on.

One more thing about the numbers, because backtests are easy to fake without meaning to. This one was not tuned to the history. The settings come from the research, not from a search for what worked. The strategy was then run through Monte Carlo simulations, parameter shifts, shuffled timing, dropped years, and out-of-sample periods it had never seen. The aim was to find out how it might behave in a less friendly world than the one on the chart. It held up. That is not a guarantee of anything; no test is. But the tests were fairly rigorous, and they were designed to break it.

Rule number one

There is a line widely attributed to Warren Buffett. Rule number one: never lose money. Rule number two: never forget rule number one.

I did not build this strategy with that in mind. I was chasing returns, and the research kept refusing to give them to me. It was only afterwards, looking at what survived, that I noticed what it had turned into: a strategy whose entire design is rule number one.

That reframed the whole thing for me. Most people’s first instinct, mine included, is to look for the fastest way up. But the arithmetic of losses is unkind. Lose half, and you need to double just to get back to even. Lose it all, and there is nothing left to compound, however good the next idea is. The people who end up wealthy are rarely the ones who found the fastest strategy. They are the ones who were still in the market, with their capital intact, when the good years came.

So I have come to think of this as a foundation rather than a strategy. It is the layer meant to keep hard-earned capital from being wiped out, so that everything built on top of it has something to stand on. Whether that view holds up is what the live track is for.

About the strategies that promise the moon

I am aware of what this is competing with for your attention. Strategies claiming triple-digit returns and Sharpe ratios in the double digits are one search away. So let me say what I think is true about them, because it is not as simple as “they’re all fake.”

Some of them are real. Returns like that do happen, and people do make good money from them. The catch is how long they last. A strategy like that tends to work for a while: weeks, months, sometimes a few years. Rarely decades. Markets change, other people find the same edge, and the thing that worked stops working. And when it stops, it rarely sends a notice. It just starts giving the money back, and it can give back a lot of it before you accept that it’s over.

So the real job is not finding one of these strategies. It is watching it every day for the moment it breaks, and having the next one ready when it does. Do that properly and it is a full-time occupation. That is why Warren Buffett has never run his money this way. It is also why the quant funds that do, Renaissance being the famous one, keep their methods secret. By every account, they run many strategies at once, with a building full of PhDs to find, monitor, and retire them.

If you try this at home, that is who you are competing against. They have more data, faster machines, and more people than you do. And if your strategy works, and it is simple enough for one person to run, there is a fair chance they will find it too. Then they will be ahead of your trades.

None of that means you shouldn’t try. It means you should know what you are signing up for. Not an early and easy retirement. A second job, with a boss who never tells you when you’re fired.

I can’t offer triple digits, and I would be suspicious of many who do. What I can offer is something rarer. A strategy whose every claim you can check, run on real money, with every miss published. One that asks nothing of you but patience.

What I would publish, if you want it

- Three tracks, always shown side by side. The backtest (hypothetical). A paper track that follows the rules exactly, started September 14 with no history behind it. And the actual account, which follows the rules as well as I can on any given day. The gap between the last two is part of the record.

- The stop rules, written down in advance. There is a pre-committed line at which I would call it broken and say so. You would know the line before it is crossed.

- How it works, in plain English. Which of the research ideas does what, and why.

- The failures. Everything that was tested and didn’t work, which is most of what was tested.

What I would not publish: trade signals, allocation targets, or anything that could be mistaken for advice. This is education and a public record, nothing more.

The question

A regular series is a real commitment, and I don’t want to make it to an empty room. So, before I decide:

If the answer is mostly “no,” that’s useful too. I’ll run it quietly, post when something is worth saying, and keep the weekly slot for the research.

Disclosures

This is educational content and a personal record, not advice. Nothing here is investment, tax, or legal advice, and nothing is a recommendation to buy, sell, or hold anything or to follow any strategy. It is not tailored to any reader’s circumstances. Reading or subscribing does not make me anyone’s adviser and creates no advisory, fiduciary, or client relationship, whatever credentials or registrations I hold now or later.

Performance comes with three labels, and each has limits. Backtested results are hypothetical, built with hindsight, and can be overfitted despite the tests described. Paper-tracked results are hypothetical too: a model following the rules with no real orders. Actual results are from my own account, net of the costs I paid, and cover a short period. All figures are approximate and unaudited. Past performance, hypothetical or actual, does not predict future results.

All investing involves risk, including loss of principal. Small past drawdowns do not mean small future ones, and markets can do what no historical record contains.

I hold the positions described, in my own accounts, and may change them at any time without notice. I receive no compensation from any issuer, broker, or platform mentioned. I have no obligation to update anything published here, and the strategy itself may change or stop.

This content is written for a general US audience, and readers act on it at their own risk and judgment.