One of the downsides of mechanical systems is their tendency to thrive in some market environments and blow up in others. If only you could tell ahead of time when a logical shut-off time was approaching, you could increase your performance substantially. Like a hibernating bear, you could take some weeks or months off, keep your powder dry and return with renewed firepower and vigor.
This concept of the trading “switch” is controversial among mechanical advocates. Some maintain that if you could design something to anticipate changing environments, you could rule the world. The implication here is that it’s not likely to happen. Others regard the idea as not so all-encompassing, but rather, more of a glorified filter.
Filters are an everyday part of mechanical systems, for example: Don’t take a given long if the RSI index is too high; don’t initiate trades if volume is lower than the lowest of the previous five day volume; if T-bonds are rising, don’t buy the stock market; etc. However, by expanding the concept you could identify complete trading environments, thereby targeting those that have promise and avoiding those that don’t.
The soybean market can exhibit one of two distinct personalities, as anyone trading it long enough can attest. You have years when the market is fairly docile and range-bound. Then, some major fundamental change occurs, usually weather related. You then see market activity that rivals coffee region freezes, oil shortages and stock market corrections in terms of volatility and opportunity.
What generalizations can you make about the hyper environment? It tends to be short-lived. On bar charts, a weather market looks like a spike with the down leg being just as violent as the up one. It’s almost certainly going to be a bull market, meaning relatively high prices. Bear markets in the grains are steady grinders. There are none of the near-instantaneous price spikes we’re talking about here. Last week’s low, for example, might have been $4.65 per bushel and this week perhaps it got down to $4.57, which is water torture for the bulls.
Almost by definition, a frenzied market will have trading ranges significantly larger than normal. Such markets need a lot of room to swing around. They have to accommodate things such a bigger than normal bid-offer spreads. It’s hard to imagine hysteria without significant price movement.
In short, we have definable threshold levels. A switch could toggle back and forth between them, on and off. Your programming code could contain something like “if the average three-day range is greater than or equal to two times the average 200-day range” for an entry qualifier. It could be something even more static like “if the close is greater than or equal to $7.00….” Seven dollar-plus beans have a distinct personality, but it’s better to trust systems more if they key off something dynamic, like expanding and contracting ranges, for example.
SOYBEAN CASE STUDY
Consider the following simple mechanical system in the soybean market: Buy tomorrow at the opening price plus twice the average three-day range on a stop; sell short at the open minus that same distance. No stops, no profit objectives. You wouldn’t want to trade it, but it does demonstrate an interesting robustness as “Bean there, done that” (below) demonstrates. Because the trades are so infrequent, we’re forced to look at a long time frame: in this case, 25 years up through the end of October 2005.
Now, let’s add the following qualifier: the highest high of last 200 days must have occurred within the last 20. In other words, although we can either buy or sell short, we’re currently in a bull market. Like before, we still exit any time that we trade two times the average range on the opposite side of the opening, but our entrances are now curtailed.
“Bulls only may apply” (below) shows the results. We see a respectable bottom line, the sum of some huge profit-pertrade figures. The bottom line vs. the worst drawdown is better than before, yielding a higher return on account, former divided by the latter. Granted, there’s some reason for suspicion. For starters, 35 total trades are hardly enough for passing judgment on theoretical performance. On the other hand, less fortification is needed if your idea is conforming to pretesting expectations.
IDENTIFYING RANGE
Momentum figures tend to be more persistent in wilder markets than inert ones. We started with a bellwether breakout idea, theorized that it would work best in an active market and got confirmation. Note that with the filter, you are in the market a mere 20% of the time instead of nearly constantly as in the first example. This is exactly what we’d expect from something that’s particular about when we should be committed.
There are many quantifiable possibilities for defining accommodating markets. Here’s another: In a trending environment, you should see more up days followed by up days and down days followed by down days. In a mish-mosh time frame, like we’ve been seeing recently in the stock indexes, more up will be followed by down and vice versa.
We give two same-directional closes in a row a “+1” value. Let’s call the resulting variable x. If it’s an up, followed by a down or vice versa, then x = 0. If over a given number of days, x is greater than or equal to a given threshold, we’ll trade.
“Capturing the trend” (below) shows a variation on the simple breakout idea in nine financial markets. This time we’re using half the value of the three-day average range as our buy and sell stop entry levels.
The top line in each market shows the basic system with no added qualifiers. The second line filter stipulates at least 60% of the last 15-day closes must be in the same direction as the previous day (x >= 0.6). Again, the qualifier affects the entry only, not the exit.
The testing period is from Jan. 3, 1999, through Oct. 28, 2005. We don’t get compliance across the board. The five-year Treasury note, Japanese yen and euro are made worse by the filter. However, note the vast improvement of the remaining six markets. Some of the actual bottom lines are bigger, despite that our average time in the market is about one-fourth of what it was before. In several areas, the drawdowns are a fraction of what they were prefilter. Note the $26,375 figure in the S&P 500 vs. the previous one of $181,800.
Similarly, some of the profit-per-trade figures have increased by many multiples. It costs significantly less money to bankroll the filter idea, but your net profits are in the same ballpark. The switch is doing its job.
What else could provide such demarcation lines? In a trending market, you’re more likely to be trading near the multi-day highs or lows than in a middle ground. Your daily bars may look the same way: more closes within some percentage of the high or low in trending markets. Volume may provide a clue or the frequency that highs or lows are taken out within a given time frame.
The financial market system we outlined could be modified so that we’re trading daily bars off of weekly signals. You could compare total up and down closes rather than noting the number of two-day like closes. There could be some sort of angle application giving you information about the speed of a move.
Again, your potential thresholds are limited only by your imagination.
While some traders may say the need for switches is an indication a system needs more work, if a system is successful a high percentage of the time and those times that it has problems are readily identifiable, it is best not to seek perfection in the system but know when to give it a rest.
Art Collins is the author of Market Beaters and When Supertraders Meet Kryptonite. This article is derived in part from a book he is currently writing about mechanical trading techniques. E-mail: artcollins@ameritech.net.
