Boxes and Channels

Livermore's pivotal point lived in his judgment: he decided where the pressure sat. The next two generations removed the judgment. Nicolas Darvas reduced the breakout to a visible container — the box — and Richard Donchian reduced it further to a formula that needs no chart at all. This module is about what gets gained, and what gets lost, when a method becomes mechanical.

Darvas: the box

Darvas was a professional dancer who traded by cable from tour stops around the world — a constraint that became his edge, since he couldn't watch the tape or take phone tips. In his telling (How I Made $2,000,000 in the Stock Market, 1960 — paraphrased here), a stock in a healthy advance moves in stacked ranges: it runs, then bounces between a ceiling and a floor like a ball in a glass box. The box's top is defined when rallies stop reaching a new high; the floor when declines stop making new lows. His rules, in paraphrase: buy only when the stock pushes decisively out through the top of the box, place a stop just below the breakout level immediately, raise the stop from box to box as new ones form on top, and let the stop — never an opinion — end the trade. He also insisted on a fundamental screen (improving earnings power in industries with a future), calling the combination "techno-fundamentalist."

See the inheritance: the box top is a pivotal point, but drawn by rule rather than felt by intuition. And the trailing box stop is Livermore's "sitting" made survivable — you sit not because you're brave, but because the exit is pre-committed and moves up behind price.

Donchian: the channel

Donchian, who had been publishing trend-following commentary since the 1930s — his widely republished "20 trading guides" first appeared in 1934, including the essential one, paraphrased: limit losses and ride profits, irrespective of all other rules — took the final step in the 1950s–70s: remove the pattern recognition entirely. His channel needs one number: the highest high and lowest low of the past N days. Price closes above the N-day high → the trend is up, be long. Below the N-day low → be short or out. No boxes to draw, no judgment about which highs "count." Any clerk — or any computer, which is the point — can compute it.

That mechanical purity is why his rule became the seed of systematic trend following. When Richard Dennis trained the Turtles in 1984 (next module), the entries they were handed were, in the Turtles' own published words:

What mechanization buys — and costs

The gain is consistency: a rule that runs identically on every market, every day, immune to mood. That's what lets the time-series-momentum evidence from Module 1 exist at all — you can only test what's defined. The cost is whipsaw: a channel rule takes every breakout, including the many that fail, and pays for the big trends with strings of small losses. Darvas kept some discretion (the fundamental screen, the quality of the box) and accepted fewer, more selective trades; Donchian's approach accepts them all and survives on exits and diversification. Neither answer is free. Modules 5–7 show the selective branch of the family; Module 4 shows the systematic one, complete with its survival math.

Assignment

Pull up any daily chart and mark yesterday's 20-day high and 20-day low — that's a Donchian channel, drawn by hand in thirty seconds. Then look left: find one place where the channel breakout caught a real trend, and one where it whipsawed. Both happened; both always happen. Write one sentence on which failure you'd personally find harder to sit through: missing the trend, or eating the whipsaws.