The slowest strategy we sell, and why that is the point

Thirty trades. Across ten companies and five years, the rule fired about thirty times, which is to say each company was bought and sold about three times in five years. Around one trade in three made money. If you are used to thinking that a strategy needs to be right most of the time, that figure looks like a failure.

It is not, because of the sizes. The losing trades were small: a cross that reversed within a few months, sold for a little less than it was bought for. The winning trades were the ones where the cross held for a year or more, and each of those earned several times what a losing trade cost. Across the whole record the money won was more than twice the money lost, which is a comfortable margin and the thing the quality bar actually measures.

A third of the trades win. The strategy is still ahead, because being wrong is cheap and being right is held for a year.

The fifty-and-two-hundred rule could be made faster. Use ten and fifty days instead, and it would catch turns sooner and trade several times as often. It would also be wrong several times as often, because most short-term turns are noise, and it would pay brokerage on every one. The slowness is not a limitation that a cleverer version would remove. It is the filter. Trends in large companies that last are the ones that move a ten-month average, and those are the only ones this rule can see.

The cost of the filter is lateness. The strategy buys some way above the bottom and sells some way below the top, every time. What it captures is the long middle of a rise, which, in the companies it trades, has historically been where most of the money was.

Each position has a stop well below the entry, twelve to eighteen percent depending on the version. People read that and assume it is there to protect profit. It is not. By the time a trend that has worked for a year reverses, the fifty-day average crossing back is the exit, and that cross usually arrives long before an eighteen percent fall from the entry would.

The stop is for the other case: a cross that was wrong from the first week, where the company turns down immediately and keeps going. Without the stop that trade waits months for the averages to uncross. With it, the trade is cut at a known cost. And because each position is sized so that hitting the stop costs the same small fraction of the account, a wide stop means a small position rather than a large loss.

Someone who already thinks in years, who would be bored by a strategy that trades every week, and who can hold a position through a bad month without touching it. The three versions share the signal and differ only in how much is at stake per trade, so the choice between them is a choice about how large a bad year you are prepared to have.

It is not for someone who wants activity, or who will find thirty trades in five years too few to believe in. That doubt is fair, and the article on backtests says why: a short record carries wide error bars. The rule's defence is not its five-year number. It is that the same rule has been doing roughly this in large companies for as long as anyone has kept records.