If you happen upon a middle-aged woman jogging down the highway wearing full armor and muttering “match the offset to the ADX,” that's me. I’m a woman in quest of the perfect Donchian channel breakout signal.

My quest began, appropriately enough, on a day of rest and reflection. As an OEX trader, I avoid trading on low-volume days, using those days to tweak settings and perform research. On a recent day, that research centered on Donchian channels.

Pegged one of the simplest of trend-following systems, Donchian channels pinpoint breakouts. Richard Donchian expounded on the system in a 1970 booklet, using the channels to employ his four-week rule. He constructed the Donchian channels with the top boundary set at the highest high for the previous four weeks and the lower boundary at the lowest low for that period.

He advocated always being in the market, covering short positions and entering long ones when price exceeded the top channel and liquidating long positions and entering short ones when price declined below the bottom channel. The default 20 setting on most Donchian channels reflects that four-week setting, with 20 days comprising 4 trading weeks.

This system does prove simple, but don’t confuse simplicity with ineffectiveness. Studies comparing various trading systems consistently identify breakout systems as those that produced the most profits. That information convinced me to take a serious look at Donchian channels.

I had no intention of constantly trading the OEX but was interested in identifying breakouts. To my confusion, when I applied the default Donchian channel 20-period, no-offset configuration to a 30-minute OEX chart, I saw no breakout signals.

30-Minute OEX Chart with Donchian Channel with Default (20,0) Setting

I scrolled back through a chart. Still no breakouts. How could a tool used to trigger breakout trades be helpful if it identified no breakouts? I experimented with the offsets, finally settling on an offset of +2. That offset translated the channel two periods to the right. Aha! Breakouts.

30-Minute OEX chart with Donchian Channel with (20,2) Setting

A quick calculation across several time periods showed some breakouts running for 5 - 10 OEX points, some far more. The quest had begun. I needed back tests to confirm that the (20,2) setting on a 30-minute OEX chart identified tradable breakouts. I had to determine an effective exit strategy.

Over the next weeks, I chose a two-month test period at random and applied various test strategies to that time period. My first test of the Donchian channels proved disappointing. In this test, I entered a trade on the first 30-minute close outside the Donchian channel and exited the trade on the first 30-minute close inside the channel. Fifty-four trades over that two-month period resulted in a total loss of 5.84 points. The maximum gain was 14.89 points and the maximum loss was 7.91 points.

A second test covered the same period but employed a different exit strategy. Using this strategy, I exited the trade when a 30-minute candle closed two points below the previous 30-minute close on a bullish trade and two points above on a bearish trade. That system resulted in 54.64 OEX points during the same time period, with a maximum gain of 35.93 points and a maximum loss of 8.54 points. I was onto something.

When running the tests, I noticed peculiarities relating to the first 30-minute period of the day. If the OEX opened outside the Donchian channel, it often ran for several points, but such trades usually proved more profitable if entered at the open. An entry at the close of that first 30-minute period lopped off several possible points before what often turned out to be a quick reversal. Conversely, if I carried a trade overnight, and the OEX opened inside the channel, it usually proved best to exit at the open.

I needed to verify this anecdotal evidence in a back test. I also decided to limit losses to four points from the entry, stopping the trade on a four-point move away from the entry rather than on a four-point close. Applying those parameters to the previously chosen time period and otherwise using the same exit strategy produced 91.34 OEX points, with the same maximum gain of 35.93 points and a maximum loss of 4 points. The maximum drawdown was 12.36 points.

Now I had the right system. My quest had succeeded. However, a troubling 52 percent of the 33 trades resulted in losses. Perhaps I could tweak my methodology, using 100-sma / 100-ema crossovers to guide entries. Bullish moving-average crossovers would mean that I would take no bearish entries, and bearish crossovers would mean no bullish entries.

That methodology resulted in far fewer trades. The maximum gain remained 35.93 points, and the maximum loss was four points, with a maximum drawdown of 8.36 points. The total gain was 75.08 points. A hefty 47 percent of the 19 trades still resulted in losses, however.

Switching to a 10-sma / 10-ema crossover and applying the same standards resulted in a total gain of 89.71 points for the period, with a maximum gain of 35.93 points and a maximum loss of 4 points. The maximum drawdown was 8.36 points, but 48 percent of the 25 trades resulted in losses.

Another test used an exit system that I later learned was a recommended methodology. I added a short-term (5,2) Donchian channel to the chart, nested inside the longer-term chart. Bullish exits occurred when the OEX closed a thirty-minute period below the short-term channel. Bearish exits occurred when the OEX closed a 30-minute period above the short-term channel.

I tested the same two-month period and found that the trailing stop method produced the largest gains over the entire period. However, the maximum gain on a single trade expanded to 41.34 points when a (5,2) channel was used to trigger exits, and 50 percent fewer trades resulted in almost the same total gain for the period, 88.64 points. Fifty-five percent of the 22 trades resulted in losses, however.

Next, I tested the system against a control period. The control period I chose was the two-month period that had recently concluded at the time of the test. In the second test period, when I applied the same standards with a trailing-stop exit that had resulted in a gain of 91.34 OEX points, I saw a loss of 0.68 points. Sixty-one percent of the 18 trades resulted in losses. Ouch!

Several modifications eventually produced modest gains for that period, but I ultimately concluded that the Donchian channel breakout system suffered from the shortcomings of all methodologies that work well in a trending market. They perform poorly in a range-bound one. Had my quest ended? No way. I just needed a way to distinguish those trades likely to perform poorly from those likely to perform well.

I had clues to help in my quest. I observed that trades tended to perform best when 21(3)3 stochastics were already pinned at overbought or oversold levels at the time the trade was triggered. That made sense. This evidence suggested that the trend was strong, and I already knew that a breakout system did not trigger a trade until the trend had been in place over a few points. That observation also hinted that ADX might be useful in determining which trades I should enter and which to avoid. I assumed that high ADX values, indicating a strong trend in place, would accompany the most profitable trades.

I assumed wrong. Initial scans have not confirmed my impression, with below-18 ADX values being associated with some of the most profitable trades. My initial scans also show that ADX values near 20 produced trades that often settled at breakeven at best. I must pursue those clues in future quests.

An example occurred with a trade triggered on October 22. ADX measured 22.46 at the time the trade triggered and stochastics lines measured 49.12 and 70.21. I concluded that this trade likely would not be an optimal trade. One type of exit produced a 1.13-point gain and the other a more satisfactory 4.61 gain. The following chart shows the path the trade took.

30-Minute OEX Chart with Donchian Channels Set at (20,2) and (5,2)

My impressions, still to be verified, include the following:

Donchian channel breakouts miss the first few points of a move, of course, but identify the big moves. These breakouts, using a 20-period channel with an offset of 2, also produce many false signals, during both trending and range-bound markets. When markets trend, however, the big gains more than offset smaller losses. Losing trades usually amounted to 45 - 55 percent of the total trades, but that percentage might be lessened as I perfect the entry and exit methodologies.

An open above or below the (20,2) Donchian channel often runs for several points, with a trade best entered at the open. However, that first 30-minute range often quickly reverses, so I might need a different exit system for such entries.

If a trade was active at the open and the first OEX price gapped inside the channel, closing the trade at the open often proved to be a wise choice.

Trades tended to perform best when the 21(3)3 stochastics were already pinned at overbought (in a bullish breakout) or oversold (in a bearish breakout) at the time a Donchian channel breakout triggered a trade, but further tests may refute this observation.

Contrary to logic, the trades may perform best when ADX measures less than 18 at the time a breakout triggers a trade. Further tests may refute this observation.

Many trades, once triggered, saw the OEX run for many points in the direction of the trade, but exit strategies still do not appear optimal. Future testing needs to focus on a SAR exit or on tightening the trailing stop when using the trailing-stop method.

Donchian channels prove simple to construct and simple to evaluate, so, perhaps, should be part of each technical trader’s toolkit. For now, I conclude that Donchian channel breakouts might best be used as adjuncts to other technical trading tools, confirming entries already suggested by those other means, but my quest continues.

The hope of capturing 90 OEX point moves in one two-month period keeps me going. What if ADX levels correlate closely with the trade’s success? What if the channels might be offset to the left, using negative numbers, during periods identified as range-bound, so no signals would be triggered until and unless a breakout proved particularly strong?

Next week, I strap my armor on again and set out to finish my quest. If you spot me jogging along, muttering about the correlation of ADX levels to the offset, you’ll know I’m planning my next back test of Donchian channel breakouts.

*Reprinted (and modified) with permission from OptionInvestor.com