Dry Powder

Executive summary

Most strategies are always at the table, paying the spread on every hand. This one is the opposite. It deploys only about 8% of its capital on average — holding a position on roughly a third of trading days and sitting in cash the rest — and commits only when a strong stock has sold off hard enough to snap back. A few days of work, then back to cash.

Because it is so rarely invested, its headline return looks modest next to buying the index. That is the wrong lens. The right question is whether the rare trades it does take are skill or luck — and here the evidence is unusually clean. We re-ran the strategy 40 times with the stock-picking replaced by random guesses; every one came in below the real strategy on a risk-adjusted basis (their results cluster near zero), while the real one won — in the test period and again in unseen later years. Measured over the time it is actually invested, the deployed capital compounds at about 11% a year.

Why “it trails the index” is the wrong scorecard

On a like-for-like footing the picture changes. Its worst drawdown over 21 years was about −13%, against roughly −55% for the S&P 500. It won on nearly two out of three trades, held each for about four days, and stepped almost entirely around the 2008 crash because it was sitting in cash when the market fell apart. Measured over time-in-market only, the deployed capital compounded at about 11% a year.

The catch is real and important: because the edge lives in being selective, you cannot simply pour more money in to make it bigger. Push it to trade more often and the extra trades are the low-quality ones — the edge thins out. Its strength and its ceiling are the same trait.

The proof it’s skill, not luck

That is the single most important slide for a skeptical portfolio manager. A high win rate can be luck; beating every randomized copy of yourself, twice, in and out of sample, is not.

How it works, in plain terms

• Only strong stocks. The name must be in an uptrend (above its long-term average).
• Only real selling. It must close at fresh short-term lows two days running, then a buy order is placed a few percent below — filled only if the selling actually continues. No panic, no fill.
• Quick exit. Sell into the bounce, or after about a week — then back to cash.

Methodology is Glenn Osborne’s “HV-RSI”, implemented faithfully and tested without curve-fitting, on split- and dividend-adjusted prices.

Tested the hard way

Real history, including the losers. The S&P 500 membership is point-in-time and includes companies that later failed or were acquired — so the result isn’t flattered by only trading today’s survivors.

It holds up on unseen data. Split the history in two and the edge is just as strong in the later half it was never tuned on (profit factor 1.55 → 1.61, win rate flat). A small-cap cross-check on the Russell 2000 is a planned next test on this data.

Costs, slippage and cash, stated plainly. We modelled Interactive Brokers commissions — they lower the return by under ~10 bps a year, immaterial. Slippage is the friction that matters, because the strategy trades its small deployed slice often and buys into dislocations; the Deep Dive bounds it. And because the book sits ~90% in cash, the headline earns 0% on that balance — a complete picture requires adding the true cash return you would earn on the idle balance at your own rate.

Who this is for

This fits a portfolio manager looking for an uncorrelated, capital-light sleeve — a genuine equity edge that runs on a thin slice of the book and leaves the rest in cash, earning yield or funding other strategies. It is a complement to a portfolio, valued for the shape of its returns (smooth, liquid, crash-resistant) rather than their raw size. As a single, standalone engine meant to compound aggressively, it is structurally under-deployed — it is built to be patient, not big.

→ Full Deep Dive with the complete methodology, the random-portfolio test, the in/out-of-sample work, and risk-adjusted tables.