Multiple timeframe analysis is well known to many discretionary day-traders as a golden principle of long-term trading success. However, it’s rare to come across solid, well-tested information on how to utilize this concept the right way. By demonstrating specific techniques of combining multiple time frame information into a single profitable model — building off the backbone of a basic volume-based breakout approach — we can make strides toward understanding how this is best accomplished.
A volume breakout on a five-minute chart is one way traders confirm a price move direction. One of the simplest forms of such a setup is to look for bars having 150% volume over the previous bar, and then hop in to grab a small profit off the move. The assumption is that the trend will continue at least for a short period of time due to the excessive interest shown based on high volume.
For the testing of this approach, the following data will be used: five-minute E-mini S&P 500 data for the 9 a.m. to 4 p.m. (Eastern) time period from Jan. 1, 2005, to Jan. 31, 2008.
The rules are as follows:
1. Look for volume of current five-minute bar > 150% of volume of previous five-minute bar
2. Long if current five-minute bar close > previous five-minute bar close
3. Short if current five-minute bar close < previous five-minute bar close
4. Exit position on a 2.5-point profit target
5. All positions exit at 4 p.m. (Eastern) if not closed by then
6. Assume commission of $5 per contract roundturn
The equity curve in “Volume breakout” shows how this system performed. At first glance, things don’t look so rosy. Indeed, 57% of the trades in this model are losers. From the surface, we have no reason to investigate any further. The equity curve looks like a great recipe for draining your account. However, when we put a little more thought into the numbers, we see that the commission paid is more than the net loss of the system. That means, with only 43% of winning trades, the model might manage to survive if we can rein in commissions. In other words, the bias may exist, just not on a level to overcome the transaction costs, at least in this basic form.
One factor that directly affects the amount of commissions we pay is the number of trades. We need to reduce them, hopefully eliminating more losing ones than winning ones. An effective approach to this is to look for simple time frame filters.
Common sense suggests that trading volume should drop appreciably during the lunch hour on most trading days. When volume is too low, the volume breakout condition could generate more losing trades than usual due to the relatively low volume we’d expect to see on the reference bar.
However, rather than applying a static filter that simply blocks trading during lunch time, a more dynamic approach is to add an additional volume reference on a longer time frame. In addition to gaining the benefit of expanding our viewpoint, this reference also can filter out signals independent of time, such as those generated on slow trading days due to regional holidays or before long weekends.
An effective reference point for this filter is the previous day’s volume. By only trading when the volume of the current five-minute bar is greater than 150% of the average volume per five-minute bar from the previous trading day, we can quantify this relationship.
“On high volume” shows the result of using this filter. The equity curve VBO Daily Filter shown in the chart has improved drastically over the original model. It is telling us that this particular filter has done something that we intended. However, this system is not there yet.
We must continue our search for better filters.
Another well-known filtering technique common for day-traders is to trade only in the direction of the moving average in a higher time frame. For example, when the price is trading above the particular moving average, traders look for long trades only. When the price is trading below that, only short side trades are taken.
Using a 20-period moving average on 30-minute chart is a common directional indicator. We can apply that as a filter to our volume breakout model to look for improvements over the basic system.
Filtering rules are as follows:
• Take long signals only if the current five-minute bar closes over the previous 30-minute 20-period moving average.
• Take short signals only if the current five-minute bar closes under the previous 30-minute 20-period moving average.
The equity curve VBO 30-minute filter in “On high volume” shows marked improvement over the original model. In terms of net profit, it is not much better than the one that employs the daily volume filter. However, this filter is better in terms of smaller drawdown. Still, it is lacking performance consistencies.
Up to this point, our research clearly shows that none of the three models is something we would like to trade on their own. They lack consistent profits, trade too much or experience extended drawdowns. We still need something that performs better.
MORE THAN THE SUM…
One solution should be readily apparent. We can combine the two filters together in the same model.
“Dual filters” shows the equity curve of the combined model. The performance statistics of both filters, tested independently, and the combined system are shown in “System performance.”
The resulting combined model is far superior to the ones with their individual filters. It has a better overall performance and a smoother equity curve. It also should be noted that none of the parameters used in this model have been optimized. The values are defaults and have not been changed to better perform on this particular data set.
The combined model shows that combining several classic day-trading concepts, even if they are traditionally employed on varying time frames, is sound. The hard part is that if you apply these concepts blindly, they can work against you. There are certain times and circumstances where these concepts work. Other times they are no better than random. By testing the concepts ourselves and then combining them with other simple filters and setups, we have gained insight into how they perform and can reliably determine whether they fit into our trading plan.
Day-trading setups and models that discretionary traders employ are different from those trading models designed from the ground up for mechanical trading. Good day-traders who trade discretionary use many trading setups where each setup has its own profitability profile.
As it stands, the combined model shown here serves as a starting point for anyone interested in using multiple time frame analysis in their trading. New rules can be added easily to the system to see what effects they have on the overall performance. The code for the system can be accessed on the Futures Web site downloads page.
The model developed here is simply a scalping setup for day trading. It suffers the common issues with many other scalping-style methods in that it has to trade frequently and that it has long flat periods where it cannot make any money. For example, the combined model makes no money at all from the middle of 2006 to early 2007. Such behavior, of course, is better than having huge drawdown from time to time.
Another aspect with a discretionary scalping model is that profit per contract is slim. Over a long period of time, this type of model (or trading setup) usually nets within a tick or two, on a per contract basis. Thus, the cost of commissions and the execution skill of the trader are critical.
Lawrence Chan is a trader and principal designer of NeoTicker. Reach him at firstname.lastname@example.org.