Most short-term breakout methodologies, such as the opening-range breakout, worked well until 2008, but then fell apart because of aspects of electronic trading. All is not lost, however. We may be able to identify a new breakout point, defined by the stretch formula, for today’s electronic markets.
Our solution will address problems associated with a reference point breakout method that keys off the open. This traditional approach has two major issues. The first is defining a stop level that takes into account enough market noise to allow for a valid breakout. This is necessary to ensure that we take a breakout in the right direction. The second issue is subsequent retracement — simply whether the breakout takes hold.
We first must define an appropriate noise level to identify valid range expansion. This allows us to test a given reference as a breakout anchor, how well a given breakout holds and the effects of different filters on the breakout. We will research this by studying the next bar’s high and low and comparing it to the current bar’s close. This current close will be the anchor for our breakout.
Our methodology measures the relevant stretch, or breakout threshold, off the current bar’s close. We will use the minimum of the next bar’s high minus the set-point, or the set-point minus the next bar’s low. Our goal is to find the set-point that produces the best profit for the research period. Next, we will identify a known factor to help us predict the direction of the breakout and the noise level of the market.
We’ll begin by using the close as the set-point. The code references daily bars, so only one breakout can be taken per day. The direction of the range expansion is known, and a breakout can be taken only in that direction.
Sub PerfectBreakout(Hold,Direct)
Dim TomorHigh As BarArray
Dim TomorLow As BarArray
TomorHigh=tsgetdatastream(0,”High”,-1)
TomorLow=tsgetdatastream(0,”Low”,-1)
If TomorHigh-Close>Close-TomorLow Then
If Direct>=0 then
Buy(“”,1,Close+(Close-TomorLow)+getactiveminmove(),Stop,Day)
End if
Else
If Direct>=0 then
Sell(“”,1,Close-(TomorHigh-Close)-getactiveminmove(),Stop,Day)
End if
End If
If barssinceentry>Hold then ExitLong(“”,””,1,0,CloseExit,Day)
If barssinceentry>Hold then ExitShort(“”,””,1,0,CloseExit,Day)
End Sub
We calculate the noise off the previous close. It is defined as the minimum move off the previous close for the next day. We then take the trade in the other direction. (We don’t know this ahead of time, but here we are studying this tendency as a research tool.) If we did not use future results, we would not be able to tell whether a losing trade was because of a breakout’s retracement, or if the breakout was in the wrong direction because of incorrect analysis of the noise level.
Noise study results
The next step is to examine the results of our study. We will conduct it on E-mini S&P 500 and E-mini Nasdaq 100 futures contracts. Again, this is a study only, not an actual trading system, because the methodology looks into the future to gather the necessary data.
Our research will help us define a better stretch formula based on the close and allow us to design better breakout filters. As a reference point, a perfect stretch is a breakout just one tick above the market movement in the other direction. We can use this model to create filters that allow us to find a breakout that moves further away from the balance point, both without retracement and with a defined exit. We also can use our analysis to model our noise factor.
Our study is conducted from Sept. 12, 1997, to Aug. 6, 2012. Results for the E-mini S&P 500 and Nasdaq 100 are shown in “Stretch studies” (below). Analysis shows that about 37% of the breakouts retrace by the end of the day below the offset plus noise factor. The average movement is a bit more than four points. The long and short sides are about even. Nasdaq results are similar except for the slight difference in the breakout, which holds above the entry point by the end of the day. If we adjust the holding period and exit at the close of the next day or even two to three days later, the dollar value in total profits does not increase. However, the average trade values increases because there are fewer average trades.
Click here for a more detailed analysis of the E-mini S&P 500 and E-mini Nasdaq 100 studies below.
We can attempt to find filters that will improve our results, but many do not work. Taking trades only during periods of high volatility does not help. Also, trading only when the range is greater than the average range over a defined period (such as 40) does not improve results. When we require extreme volatility, although the Nasdaq results improve some, we have too few trades to be confident in the results. Another non-performing filter is a simple counter-trend filter, such as buying only when the close is lower than some previous close or selling when the close is higher than some previous close.
Point of reference
Although our search for a reliable filter came up short, this analysis helps us understand how reference point breakouts work. We can understand what a given filter accomplishes based on how it performs with both the perfect breakout test and a classic range-based reference point breakout.
The electronic markets have destroyed opening range breakouts for futures markets. Previously, markets were closed for 12 hours or more ,whereas now they are closed only for a few hours, if that (30 minutes for E-minis). This has, in turn, made the open ineffective. Although the close is a valuable reference point, it never worked as well as the open, which was a powerful reflection of overnight sentiment shifts because of various reports, such as inflation and unemployment. While the open has lost its significance for commodity futures, stocks still are sensitive to this price point. Although most stocks trade overnight, volume is light and for N.Y. Stock Exchange stocks, we still have a specialist involved in opening the stocks each day.
Almost all day-trading breakout methods exhibited peak equity in 2008. Using what we have learned, we can develop patterns that have performed more consistently over time in electronic markets. Our first pattern will test buys on daily bars so that we can take breakouts only in a predefined direction. We will look for profits before and after 2008 and for a strategy that works for both the E-mini S&P and the Nasdaq.
Here’s the system:
Sub AnotherBreakout(LB)
Dim BalancePoint As BarArray
Dim Mom As BarArray
Mom=Close-Average(Close,LB,0)
If Mom<0 Then
Buy(“”,1,Close+Average(Max(Close[1]-Low,High-Close[1]),3,0),Stop,Day)
End If
ExitLong(“”,””,1,0,CloseExit,Day)
End Sub
We simply buy on a breakout tomorrow at the maximum noise level today as discovered from our earlier study for when prices are below a given moving average. When we optimized this moving average length, we found that the parameters were robust over a large range. We selected 40 as the test case. The results are shown in “Simple breakout: S&P 500” (below).
This pattern does better than other breakout methodologies we have considered for electronic markets. It works well on the E-mini S&P 500. The drawdown is low. We need to conduct further tests to determine if we can trade this strategy, but as a first step toward building a breakout system that will work in today’s markets, this is a good one.
Performance in the E-mini Nasdaq 100 is shown in “Simple breakout: Nasdaq” (below). The results are not as good as those for the E-mini S&P 500; nonetheless, they are better than classic opening-range breakout methodology.
When we consider the mirror image pattern on the short side, we get poor results. This is interesting because the period tested included significant stretches of both strength and weakness, which should eliminate an underlying market bias toward long trades. The initial conclusion is that we simply found a pattern that worked well for the long side but not the short side.
The electronic markets have changed the day-trading landscape and made the opening-range breakout null and void. However, by developing new reference points, we have furthered the relevance of this venerable technique in stock-based markets. Additional improvement could be had by employing neural networks to predict, for example, the direction of the range expansion and the noise level so we could evolve our approach on the fly. Machine-based rule induction also could be used to identify new concepts of breakouts and create rules to predict best breakouts, minimum retracements, direction of range expansion and the noise level.
Murray A. Ruggiero Jr. is the author of “Cybernetic Trading Strategies” (Wiley). E-mail him at ruggieroassoc@aol.com.
