The Channel Named After Donchian Does Not Appear in His Own 20 Guidelines — and the Channel Leaves Today’s Bar Out of Its Own Calculation

Source 4 presents “20 trading guidelines developed by Richard Donchian, the father of trend following.” — “First published in 1934, many of the 20 trading guidelines from Richard Donchian are as relevant today as they were during the golden age of technical analysis.” They are “Based on his experiences, Donchian developed 20 trading guidelines split into two groups: general and technical.”^[4]

We searched that page for the vocabulary of the indicator that bears his name. The words “channel”, “highest high” and “lowest low” appear zero times. The guidelines are about trend lines, volume, triangles, gaps and consolidations — for instance “Reversal or resistance to a move is likely to be encountered upon reaching levels at which, in the past, the commodity has fluctuated for a considerable length of time within a narrow range”^[4]

And the system he is credited with building first was not a channel either: “Donchian developed one of the first trend-following systems based on two different moving averages, which were cutting edge in the early thirties.”^[4]

So the indicator's name records an attribution, not a citation. This page does not claim Donchian never used such a channel — only that it is absent from the 20 guidelines as this source presents them, and that “The guidelines shown below have been paraphrased for a more straightforward explanation.” with “A version of Donchian's original guidelines can be found on the Trading Tribe website (www.seykota.com).”^[4]

Source 3 gives the construction plainly: “The upper channel is set at the x-period high and the lower channel is set at the x-period low.” “The dotted centerline is the midpoint between the two channel lines.” with “Centerline: (20-day high + 20-day low)/2” and notes “This indicator, developed by Richard Donchian, is sometimes referred to as Donchian Channels.”^[3]

Then a detail the formula as usually written leaves out: “The Price Channel formula doesn't include the most recent period. Price Channels are based on prices prior to the current period.” “A 20-day Price Channel for October 21 would be based on the 20-day high and 20-day low ending the day before, October 20.”^[3]

And the reason, which is a matter of logic rather than preference: “A channel break would not be possible if the most recent period was used.” If today's high were inside the window, today's high could never exceed it — the signal would be impossible by construction.^[3]

This is exactly the kind of rule that separates implementations, and it is the sort of thing a definition alone does not tell you. “Shorter look-back periods, such as 10 days, produce tighter channel lines. Longer look-back periods produce wider channels.”^[3]

Source 5 is another practitioner's account of the same family. “Along the way we got another fine example of envelopes, Donchian bands, which consist of the highest high and lowest low of the immediately prior n-days.” Its definition matches sources 1 and 3 — three independent statements of the same construction.^[5]

And it places the channel by what makes it adapt: “Percentage bands are fixed, they do not adapt to changing market conditions; Donchian bands use recent highs and lows and Keltner bands use Average True Range as adaptive mechanisms.” — “Today the most popular approaches to trading bands are Donchian, Keltner, Percentage and, of course, Bollinger Bands.”^[5]

Source 3 adds that the horizon changes what the break means: “A move above the 20-day Price Channel signals a new 20-day high. A move above the 20-week Price Channel signals a new 20-week high. A 20-week high is more consequential than a 20-day high.”^[3]

On its own worked example: “These channel breaks caught a few good trends, but there were two whipsaws or bad signals. Indicator signals are not perfect, and there will be whipsaws.” “It's part of the game.”^[3]

Its suggested filter is a change of data, not a change of rule: “To filter signals further, you could use a close-only line plot.”^[3]

This page takes no position on whether channel breakouts are profitable and carries no backtest of its own.^[3]

The construction is the simplest of the common indicators. The channel "is formed by taking the highest high and the lowest low of the last n periods", so Upper = HighestHigh(N), Lower = LowestLow(N) and Middle = (Upper + Lower) / 2. There is no averaging and no standard deviation — which means none of the divisor or smoothing ambiguities that affect Bollinger Bands or RSI arise here. It was developed by Richard Donchian.^[1]

The stated signal rule is symmetric. "If a security trades above its highest n periods high, then a long is established. If it trades below its lowest n periods low, then a short is established." The channel also doubles as a volatility picture: "If a price is stable the Donchian channel will be relatively narrow. If the price fluctuates a lot the Donchian channel will be wider." Originally the periods were daily, though the article notes modern platforms allow any interval.^[1]

The article does not stop at the definition. It states that "extensive backtesting utilizing 360 years of exchange data and 4,887 test trades with the default setting of 20 indicates that the Donchian channel indicator was ineffective." The reported win rate is 35% of trades.^[1]

The arithmetic that follows is worth reading carefully, because it is nearly positive and then is not. Combined with a reward-to-risk ratio of 2.0:1, that win rate "yields an expected profit of 5 cents per dollar invested." "While this is positive, when combined with slippage, this indicator might be unprofitable." So the edge, as the source describes it, is small enough that execution costs can erase it — not a conclusion this page reached, but the one the reference states. This page did not verify the backtest or locate its underlying study.^[1]

The best-known application of a channel-breakout rule came from a wager. Richard Dennis "believed that successful trading could be taught", and "to settle a debate on that point with William Eckhardt" he "recruited and trained 21 men and two women, in two groups, one from December 1983, and the other from December 1984." The group became known as the Turtles.^[2]

The training was brief and the rule was a range breakout. Dennis trained them "for only two weeks about a simple trend-following system, trading a range of commodities, currencies, and bond markets, buying when prices increased above their recent range, and selling when they fell below their recent range." They were "taught to cut position size during losing periods and to pyramid aggressively." Position sizing was part of the system, not an afterthought. After a one-month trial capped at 12 contracts per market, successful participants managed $250,000 to $2 million of Dennis's own funds, and the experiment "reportedly had earned an aggregate profit of $175 million" when it ended five years later.^[2]

This is the part usually left out of the Turtle story. The article notes the exact system "has been published in at least two books and can be back-tested to check its performance in recent years", and reports the outcome plainly: "The result of such back-test shows a drastic drop in performance after 1986, and even a flat performance from 1996 to 2009." It adds that several Turtles nonetheless went on to run their own firms.^[2]

The article also characterises the era the original results came from. Describing the trending inflationary markets of the 1970s — repeated crop failures and the 1972 Soviet wheat purchases — it says this was a period in which "anyone with a simple trend-following method and a dart board could make a million dollars." Read together, the source is pointing at the market regime rather than the rule. This page does not attempt to apportion the result between the two, and notes that Dennis himself "reportedly lost $10 million" in the 1987 crash and settled investor complaints of failing to follow his own rules for over $2.5 million in 1990, without admitting or denying wrongdoing.^[2]

This page now rests on five sources across three institutions: two encyclopedia articles, two pages of a charting platform's documentation, and one practitioner's account. The previous version rested on en.wikipedia.org alone.^{[1][2][3][4][5]}

Source 4 is a paraphrase, and says so. It presents Donchian's guidelines in reworded form and points to a version on a third-party site as the original. So the absence of “channel” is an absence from this presentation of the guidelines, not proof about every text Donchian wrote.^[4]

Vendor and practitioner limitations: source 3 is StockCharts describing the overlay it implements, so its current-bar exclusion is that platform's rule; source 5 is John Bollinger describing a competing family of bands, which makes it a useful outside description but not a neutral one.^[3][5]

Not established: what “360 years of exchange data” in source 1 covers; whether any channel breakout rule is profitable; what the Turtles' exact published rules were, since only secondary accounts were read. No backtest figures of this page's own appear anywhere on it.^[1][2]

Retrieved 2026-09-28. Sources 3 and 4 HTTP 200 and read in full (133 and 140 lines); 30 quotations across the three new sources were each searched for in the fetched text, and the absence of “channel”, “highest high” and “lowest low” from source 4 was confirmed by counting those expressions themselves.^[3][4]