Saturday, September 10, 2011

Smackdown: Joe Ross vs. the Turtles

For most of the past year, Private Trader has documented my analysis -- and trades -- using a variant of the Turtle trading method popularized by Curtis Faith in his book Way of the Turtle: The Secret Methods that Turned Ordinary People into Legendary Traders. The method itself was developed by Richard Dennis, a commodities trader once known as the Prince of the Pit.

The method has been OK. It has produced some profitable trades and warned me away from some really bad positions.

But it has flaws, as I wrote in a recent essay, "The Trouble With Turtles". As a fact-based trader, when my trading tools need a tune-up, I'm under the hood in a flash.

Turtle trading's main strength lies in the fact that the analysis is primarily based on current price. No moving averages. Getting in to a position requires a breakout beyond a defined price level.

Current price carries an advantage, because averages-based tools have a built-in lag. If you're using a 20-day moving average, then you're always as much as a month behind the present, particularly if there was a sharp price movement at some point in the period.

The backward-looking averages contradict the nature of trading, which is first and foremost a forward-looking exercise: Where will prices be tomorrow? Next week? Next year? Don't waste my time with yesterdays. Speak to me of today.

So in seeking improvements to my trading methods, I don't want to throw out the price-based nature of my tools. I must, instead, look to other price-based systems for ideas on how to make things better.

Commodity trader Joe Ross has a reputation as one of the best technical traders around. If the price of his books is any indication (always three figures), he has to be at the top of his game.

Ross' method, documented in Trading by the Book, is thoroughly based on current price, and yet with significant differences from the Turtle method. I think it's a good place to look for ideas.

I'll proceed with 1) A summary of each method, 2) A comparison of the methods and dicussion of the relative advantage each has to offer, and 3) a synthesis using elements of each method to produce a set of trading rules that, I hope, will improve my results.

And then there's Step 4): Implement the synthesized method in real trading, the only way to truly understand how a system performs under pressure.

1) Summary of the Methods

Turtle trading uses Donchian price channels as the framework of analysis. A Donchian channel (named after the commodities trader Richard Donchian) draws a horizontal line at the highest and lowest prices within the past whatever number of days.

Classic Turtle trading sets 55 days as the time horizon. So, imagine a six-month stock chart with a channel as wide as the highest price and lowest price for the past 55 days. A stock trading within the channel is considered to be in neutral phase. A breakout beyond the channel is a signal to open a position, bullish for a breakout upward, bearish for a breakout downward.

But not all breakouts result in trades. In order to ensure that the price is indeed trending, the Turtle rule requires that the 14-day average directional index (adx) be increasing. The adx rises when a stock is trending regardless of the trend's direction.

In addition to the classical rules, I've some added requirements. For stocks, I require that the adx be at 25 or higher, signifying a strong trend. I want the parabolic sar to be aligned with the direction of the breakout. I want the day's price movement at the time I open a position to be in line with the breakout -- a rising price for a bullish breakout, and a falling price for a bearish one. And I want the price at the time I enter to be outside of the channel. This is to avoid breakouts that immediately pull back to within the channel.

Turtle trading sets no exit target. It rides a stock for as long as the trend continues.

For getting out of positions, classic Turtle trading adjusts for a stock's volatility, using a measure called the 14-day average true range (atr). A trailing stop/loss is set at twice the atr below the peak price for a bullish breakout, or above the lowest price for a bearish breakout.

I've added a few modifications to the exit rules. I'll uses the single atr (not doubled) or possibly half the atr for a stop/loss when trading currencies, because of the high leverage. I exit a position if the price closes within the price channel two days in a row after breakout day. I also exit a position if the parabolic sar changes against the position.

You'll notice some magic numbers in the above discussion, "magic" because who knows why they've been chosen. The 55-day period is arbitrary. So is the 14 days for the adx and the atr. I use a standard parabolic sar, with an accelerator factor 0.02 and an acceleration limit of 0.2. I have no idea why the parabolic sar's inventor, J. Welles Wilder Jr., was so obsessed with the number 2.

(Wilder also invented the average directional index and the average true range. He truly is the Thomas Edison of technical trading tools.)

Next, I turn to Joe Ross' breakout trading methods. Watch for his magic numbers.

Ross is a great trader. His books are filled with charts and anecdotes. I love reading them. But when I ask, "How can I do that?", things get a bit more complicated. For Ross doesn't write succinctly. It is not his way to set things down in a list of rules. To implement Ross requires a lot of searching, and the reader can never truly know whether he or she has gotten it right.

So my description of Joe Ross' ideas will, of necessity, be something of an impression. Call it Rossish or Ross-like trading. If you want to know precisely WWJD -- What would Joe do -- don't ask me. Buy the book and try to work it out for yourself. Perhaps you can succeed better than I.

Ross trades stocks whose prices have broken free of an area of congestion, a period when the stock has traded sideways within a well defined range for a period of time. Ross likes to see congestion lasting at least 22 bars (days on a daily chart, weeks on a weekly chart, 30-minute periods on a half-hour chart).

The channel is defined by the high price and the low price of the congestion period.

Ross then defines breakout lines by by subtracting the low price from the high price, and then multiplying the difference by 0.236 (23.6%). The result is the cushion that Ross uses to set his breakout points.

The upper breakout line is the high price plus the cushion. The lower line is the low price minus the cushion.

But Ross doesn't trade immediately upon breakout. He waits for the price to pull back, preferably by 38.2% or 61.8%. At that point, as the price reverses to the direction of the breakout, Ross enters the position.

Ross sets an exit target: The breakout level plus the width of the congestion for bull postions, minus the width for bear positions. A reminder, the width is the high price of the congestion minus the low price of the congestion. You can also get it by dividing the cushion by 0.236.

He uses a stop/loss that trails the extreme price (highest high for bull trades, lowest low for bear trades) by the amount of the cushion +/- 1¢.

Ross' magic numbers have a theory behind them, unlike those of the Turtle traders. They are Fibonacci ratios, levels that stocks retrace to time and time again.

To brush up on Fibonacci retracements ("Fibs", I call them, although they rarely lie), check out these articles in Wikipedia, Investopedia and Stockcharts.

Fibs are a key component of Elliott Wave analysis, as practiced by Robert Prechter Jr. I'm a huge fan of his, despite his regrettable tendency to refer to gold as "real money". See my essay, "Gold: A letter to a friend".

Prechter outlines his theories in his book, Elliott Wave Principle: Key To Market Behavior, and applies them in a forecast of what is happening now with the Great Recession in his book, Conquer the Crash: You Can Survive and Prosper in a Deflationary Depression.

So Ross makes extensive use of Fibs, and the trader using Ross-like methods should also be aware of Fib retracement levels.

2) Comparing the Methods

Both methods are aggressively based on the current price. Turtle, however, relates the current price to an arbitrary level. Ross relates the price to recent price movements, to a structure that's inherent in the price history.

One illustration of just how arbitrary Turtle's 55-day number is: In most instances, in the direction of a breakout, the 55-day price channel boundary and the 20-day boundary are identical.

Turtle also uses more magic numbers: 14-day adx, 14-day atr, a bunch of 2s parabolic sar -- all of them based on tradition, on design decision made by the tools' inventor for use in an entirely different trading and technical environment.

Ross' magic numbers, at least, are confined to a single system -- Fibonacci retracements -- that works -- I've seen it work under a variety of circumstances this year. The 22-day congestion number is arbitrary, but it's also flexible. Ross doesn't make it into a firm rule.

I give the grounded-in-reality points to Ross. My goal in setting up trading systems is to eliminate the arbitrary wherever possible. Arbitrary settings are an invitation to faith-based trading. I'm a fact-based trader.

Turtle is easy to see on the chart. I can spot a breakout in under one second. That's why I find it easy to scan 600 charts a day looking for top prospects for further analysis.

Ross requires more work to identify breakouts. And frankly, the whole concept of congestion is ill defined. How much fluctuation is too much to meet the definition? How many times within the 22 days must the price touch the high and low points to qualify as channeling? What if the fluctuations start wide and then narrow, as with a wedge pattern or an ascending or descending triangle?

With Ross, it can take me a minute or two to decide whether the congestion area is valid. So Turtle gets the ease-of-use points.

Ross is much more cautious than Turtle when it comes to entering a position. The Ross breakout lines are almost always beyond the 55-day Donchian price channel used by Turtle. And Ross' requirement for a retracement before entry adds another level of caution. The Turtle method is Breakout! Wham!! Bam!!! Bif!!! Place the trade.

I also have the impression -- I've not done a full study -- that the Ross stop/loss is tighter than that produced by the average true range method under Turtle.

Ross' retracement rule for entry reduces the chances of a whipsaw, although the tighter stop/loss increases the chance. But a tight stop/loss -- stops losses, so a whipsaw in that case causes little harm. I'll give caution points to Ross.

Finally, Ross sets exit targets, and Turtle doesn't. In this case, I give points to Turtle. If a trend is continuing, why jump off? Just ensure that your stop/loss is in place.

The only time I consider jumping in the midst of a trend is to avoid an earnings announcement, which can have incalculable impact on the price.

3) Synthesizing a System

Ross leads on points, but not in the area of ease of use. For me, that's close to being a deal killer.

I don't want to give up my 600 charts. That daily scan keeps me grounded in the pulse of the market. It gives me greater diversification. It leads me to opportunities that I would miss were I looking at a smaller universe.

With Turtle, I can crank up some music -- the Decembrists or my cousin Samantha Crain or my collection of two dozen covers of Dylan's "All Along the Watchtower" -- and whip through 600 stocks in an hour, identifying breakouts in no time at all.

So the challenge is how to bring that ease of use to Ross in order to benefit from Rossish caution and fact-grounded construction.

I think I have a solution.

A flat 20-day Donchian price channel boundary -- one that traces a horizontal line on the chart -- gives a reliable early warning of what might be a Ross breakout in the making.

When the price breaks out of the Turtle-style channel, my task would be to confirm that the price is indeed in congestion, calculate the Ross breakout levels, and put the stock on a watchlist, keeping an eye out for Ross breakout and retracement.

The use of the average directional index, the average true range and the parablic sar would be discontinued.

Although, on second thought, the average directional index (adx) could have its uses. With Ross-like trading, I don't want a trend in recent days. I want a low adx. So it might be useful in identifying congestion to look for an adx of under 25 or under 20, and shortening the adx period to 10 days. No decision on that yet. It's just a thought.

In fact, the Rossish method I've outlined isn't wildly different from Turtle trading. The check-it-out level is the same -- a break through the 20-day price channel, which in most cases is identical to the 55-day I'm now using. And Ross sets a tighter stop/loss.

I would describe Ross-like trading as Turtle trading plus wait-a-sec'.

Confession: I've used Ross on some recent intra-day currency trades, and made good profits.

This coming week, I'll be working on Step 4): Implementing the new system, including writing a set of formal rules for the How I Trade section. Until I'm ready to implement, probably on Sept. 18, I'll continue with my present Turtle trading, but with an eye out for Ross opportunities.


  1. Great article. Am trying to find a trend following system to automate for my forex trading now. Have you been able to find an online scanner which updates 55 day highs on the general market? Would be interested to find that one.

  2. My brokerage's trading platform, ThinkOrSwim, has some of the best scanning routines in the business, but sadly, none for a 55-day high or low breakout. I've not looked at other platforms.

    It could be that there's a way to construct such a scan using the programming tools that ToS has, but I've not tried that yet.

    For my price-channel trading, my procedure has been to do an automated screen for the average directional index, and then run through the results by hand. It's a fairly fast process, since Donchian channel breakouts are so easy to spot.