As traders, we continually search for the Philosopher’s Stone to help us unlock the secrets of the markets. Numerous technical tools have been developed, many of which claim to be the solution we all seek. But, do such tools actually stand up to rigorous analysis?
Traders often view things through green glasses. Sadly, Oz isn’t always made of emeralds. It’s important to shed those spectacles and subject our technical tools to the harsh realities of a statistical magnifying glass. Here, we present a method of doing just that in the context of identifying winning trades. As mentioned in prior articles, a winning trade is one with a positive expected value (see "Moving averages provide a simple solution," April 2011).
One popular trading tool is known as stochastics, or the stochastic oscillator, developed by George Lane in the late 1950s. According to Lane, it is a momentum indicator showing the location of the closing price relative to its high/low range over a fixed number of periods. Theoretically, because the oscillator is range bound, it is useful for identifying overbought and oversold levels. Lane developed the fast oscillator %D and %K and sought divergences. To reduce the choppiness of these, a slow stochastic oscillator was developed based upon a three-period simple moving average of Lane’s original indicators. Many traders swear by stochastics, but should they?
The stochastics indicator is plotted on a scale from 1 to 100. If the result is 70%, then traders get a warning indication of overbought conditions. On the other hand, if the result is 30%, then traders get a warning indication of oversold conditions.
Three types of trading rules often are applied to the stochastic oscillator. First, a buy signal typically is generated by a value under a set level, say 10% or 20%; a sell for values above, say, 90% or 80%. Second, a cross of %K above %D is considered a buy, while a cross below is considered a sell. Third, price and indicator divergences (for example, a lower low in price occurring concurrently with a higher low in stochastics) are assumed to precede turning points in price. (For more in the discretionary interpretation of stochastics, as well as the formula for its calculation, see "Trading stocks with stochastics".)
The lone application of the first rule generally is disregarded in a systematic sense. Research has shown overbought readings are not necessarily bearish. Securities can become overbought and remain so during a strong uptrend. Closing levels that are consistently near the top of the range indicate sustained buying pressure. Similarly, oversold readings are not necessarily bullish. Securities also can become oversold while the stochastic oscillator remains less than 10 during a strong downtrend. Closing levels consistently near the bottom of the range indicate sustained selling pressure.
The second rule is extremely attractive to the eye. A trader looks at a chart and sees a strong uptrend took place and, sure enough, the stochastic oscillator gave a clear buy signal near the start of the move. %K moved from beneath %D, to cross over it just as the trade erupted. What the trader often overlooks is all the times that situation took place in the oscillator, but no profit followed. This is a typical mistake that novice traders frequently make.
We are perennial optimists and generally only see those things that favor the result we wish to achieve. Many systems suffer from this. Advertisements always show us the "best case" and tend to gloss over the less sterling times. To be successful, we must apply an unbiased, rigorous test and then accept its results.
To do so, let us develop a simple signal system combining the philosophies of the first and second rules. We shall require both %D and %K to read less than 25 immediately prior to or at the point of crossing, and we shall buy when %K crosses above %D. We remain in the trade provided %K is greater than %D, but shall immediately exit on any failure. Conversely, we shall sell short if and only if %K and %D are both above 75 and %K crosses beneath %D, exiting upon a termination of that condition.
By using the 25/75 levels, we filter out all of the crossings that occur in the mid-range. This should help eliminate whipsaws. We could expand this filter to 20/80 or 15/85. Doing so might increase the win/loss ratio, but will reduce the number of trades. These levels easily can be tested by the individual trader and applied based upon personal risk/reward tolerance.
Now, we select a random commodity and test a randomly chosen period for enough times to have statistical significance — at least 30. It is essential for the fairness of the test that these variables be randomly chosen. We don’t want to fall into the trap of picking the commodity or period where we already know things will be rosy. Consequently, we randomly choose gold, 60-minute bars, starting Jan. 18, 2011. The test generates 55 trades.
Gold begins at $1,365.50 per ounce at the onset of our test and ends at $1,397.90 on Feb. 22, 2011. If a trader simply bought and held, a profit of $32.30 per ounce would have been realized. That is our benchmark for the test. However, because nobody is possessed of perfect trading ability, we shall deem achieving 75% of the ideal $32.30, or $24.22, to be a trading success. During our test period, gold trades in a range from $1,309.30 to $1,410.10 (see "Gold swings"). Given this $100 range, we have every reason to hope for success with the stochastic oscillator because such conditions generally provide lots of room for swing-type movement.
Sadly, the results do not reward us (see "Opportunity lost"). In 55 trades, we lose $21.30 before slippage and commissions. Our win/loss ratio is only 34.55%. This is dismal. At $100 per ounce, and adding a conservative $25 for slippage and commissions, we lose $3,505 during the same period a buy and hold strategy would have netted more than $3,200 profit. It is absolutely the case the oscillator captured gains during the period. Four of the 19 winning trades exceeded $950. But that represented a mere 7% of all of the trades signaled. It is far too tempting to see only these four strong winners and ignore the losing and mediocre trades.
Improving the breed
To improve these results, we shall add the third rule as a filter. We will execute a trade only if there is a divergence from the signal immediately preceding the current one. In other words, we will require two sell signals in a row to sell, and the second must have a lower %D and %K and a higher price than the preceding signal.
This filter does improve things. We only would take 14 trades of the 55 generated by the first two rules, with six of those trades producing winning results, two producing trivial gains (less than slippage), and six losing money (see "Filtered results"). We realize a net of $1,830 after slippage and commissions and raise our win/loss ratio to 50% (ignoring the two trivial trades). Although this still is far below the buy and hold profit of $3,200, or even the success threshold of $2,422, it is still much better than the $3,505 loss we would have experienced before application of the divergence filter.
Nevertheless, we must be careful of putting too much credence in these results. A sample size of 14 is not considered statistically significant. Additional tests on other markets, chosen at random, should be performed to determine if all three rules combined could produce similar results over a larger sample size. It also may be possible to improve the results further by application of stops, profit targets or other unrelated filters.
A stochastic oscillator is a useful tool, but has significant limitations. It will capture the large gains, but at the expense of many losing trades. The losers can be reduced to an extent by application of several filters. When analyzing any technical or fundamental trading tool, it is vital to apply a rigorous, unbiased test.
It is all too easy to see only the glitter of the Emerald City and not notice the tarnish. No single trading tool is the answer, or we all would be successful traders. The stochastic oscillator may be best utilized to see periods of potential trend change. However, the astute trader will be cautious of any other standalone tool, and should combine separate tools logically to see the market in its proper context.
Arthur Field has a Ph.D. from Clemson and is a former fund manager for Fidelity International. He wrote "The Magic 8: The Only 8 Indicators You Need to Make Millions in the Markets," available at www.themagic8.com. Email him at firstname.lastname@example.org.