This website uses cookies
Read our Privacy policy and Terms of use for more information.
That rule is the whole method. We publish every part of it, and nobody overrules it by hand, including me.
Each model reads the market a different way, and each has conditions it handles badly. Requiring three of the four keeps one model’s blind spot out of the portfolio.
The name’s own price, with bonds, the dollar, and credit around it.
Falls shortWhen a move runs further than history allows. It then turns cautious too early.
The same evidence, learned a different way, through recurring shapes over short and long spans.
Falls shortWhen that evidence misleads, because it shares it with the machine-learning model.
Price, and nothing else.
Falls shortAt turning points, where it is always late, and in choppy markets that go nowhere.
The state of the whole market: risk-on, risk-off, or neither.
Falls shortOn single names. It can read the season right and one company wrong.
A name enters the portfolio only when its own signal is long and at least three of the four models agree. It leaves when either test fails: its signal turns flat, or fewer than three models agree. The money then waits in cash.
Cash is a position here, not a leftover. The portfolio is long only, never levered, and scaled to a volatility target.
We publish the list of names. Each name gets its own line-up: the models whose reading sets its signal, long or flat, chosen by what did best over its own history.
The four models read every name. We lock each reading the moment it sends, and it stays readable afterwards.
Up to ten names that pass the rule become the portfolio. We size each by its risk and cap it, so no single name decides the year.
The sell check runs before the open. A holding leaves when its signal turns flat or agreement drops below three of four. Smaller weight changes wait for Monday.
The first Macro & Megacaps portfolio held ten names and kept 20.9% in cash. Three names sat at the 15% cap.
Each weight as a share of the invested 79.1%.
When fewer names pass the rule, the portfolio holds more cash. Expect stretches where a large share sits in cash by design.
Ten holdings at most, and often fewer. Macro & Megacaps averaged 8.4 in the backtest.
The backtest takes out trading costs, about 0.9% a year for Macro & Megacaps. It does not model tax, which in a taxable account is real.
We publish the portfolio as research: tickers and weights. It is not a managed account and not advice.
Three of four agreeing is a high bar, not a guarantee. There will be months when the models are wrong together.
The portfolios began on 3 August 2026. Everything before that date is backtest, and we label it that way everywhere.
Five notes set out every part of it. All are free to read, and they read best in this order.
You get each morning’s portfolio, and the Multi-model Signals behind it, from the first day. Cancel before the 30 days end and you pay nothing.