Hi, I've been looking at some of the work that John Ehlers has done, and I've fooled around a bit with his code to create some adaptive moving averages that vary their length to that of the dominant market cycle length. I've also dug into non-linear equations (very heady stuff!), and figured out how to create a volume-weighted, momentum-based prediction algorithm based on some of Ehler's code, as well. And I created a 'trade squelcher' that filters out potentially smaller trades, to prevent getting whipsawed so much, again based in large part on Ehler's code. What I wanted to know is...before I go galavanting off to regions unknown in search of heretofore unheard of indicators, does the Ehler's cycle analysis code and his adaptive M.A. (Specifically MAMA/FAMA) work? Is anyone actually profitably trading using these indicators? I know my backtesting tells me they should work, but I don't put much credence in backtesting, I prefer to hear from actual traders. Thanks much for any info, DarrenBrothers@yahoo.com
DGB, I spent 2 years working full time on a few sets of proprietary adaptive tools, one set similar to those you are considering, one not. While I cannot comment on Ehler's work, a quick one on my own: The results were very good, often remarkably so, but in the final analysis the very nature of the tools, specifically their adaptability to market movement, meant they would not fit my risk/reward targets for trading. For example, in an explosive move, the tools may work very well at describing market behavior, but because the tools themselves are responding to these new market conditions, structuring trades and controlling risk can be difficult. Still, the idea of adaptive tools is attractive, some findings in behavioral economics certainly suggest their usefulness, and perhaps you will have an insight I did not. I would be interested in hearing whether you make progress and decide to incorporate them in your trading. Good luck! O.
Hi, One... Ok, I'm done with the coding for my indicators, it wasn't as hard as I thought it would be. Turns out EasyLanguage is just a derivative of Pascal, and I already know that programming language, so picking it up was a piece of cake. What I did was rework Ehler's code to tighten up the dominant market cycle length calculation algorithms. Then I reworked some traditional indicators so they automatically adjusted their length as the dominant market cycle length changed. We know in signal analysis that in order to effectively sample the signal, the Nyquist frequency must be at least half that of the dominant frequency. In other words, your indicator lengths must be half the length of the dominant market cycle length. So, I jiggered with the code on the following indicators: DMI+ DMI- ADX Triple SMA (3SMA) Bollinger bands And I created my own Signal-To-Noise Ratio (SNR) indicator (I didn't like the one Ehlers created, it didn't seem to work very well). So, my whole premise is to only trade when the SNR is higher than a certain setpoint (this keeps you away from the chop). You wait for the modified Bollinger bands to show a channel that is narrow, and the ADX crosses over the lower of either DMI+ or DMI-. If DMI+ is higher than DMI-, you know to go long; vice versa for DMI- being higher than DMI+. The Triple Simple Moving Average is there as a confirmation/visual aid (I modified it so that if the trade weakens, all three lines converge into one and their length shortens to 1 bar (so it tracks the market very closely, basically to get it out of the way during non-trending markets, because I want to try to develop a consolidation phase trading scheme as well, and with the 3SMA in the way, it's hard to see the price bars in the consolidation phase), so if the 3SMA does converge and shorten its length, I know the trend is getting too weak to be viable). I exit when either the SNR drops below the threshold, or the ADX drops below both the DMI+ and DMI- lines (program exit). I also have a 'trailing stop exit' set at 10 points, and a 'stale trade' stop that automatically gets you out if the trade doesn't develop as planned and just languishes. Most trades are exited just prior to the trend halting, using the program exits. A few were stale trade exits, and only a couple hit the trailing stops (giving back 10 points from the profits). I've only done informal walk-forward testing so far... the largest single trade drawdown is 10 points (due to a price shock on a report day, I'll be programming in the report days so all open positions are covered with options or exited on report days, depending upon trend strength), the largest single trade updraw is 84 points over 2 1/2 days. Average is 18-20 points per trend. I used 2 points round-turn as slippage, and 1 point as commissions. This is all done on a 5 minute chart. So far, it seems to work quite well. The questions I have for you is: 1.) What specific problems did you have in regards to the indicators readjusting themselves to market parameters? 2.) How did you overcome these problems? 3.) Did you scale into or out of positions, and if so, what method did you use to determine the number of contracts for each scale? 4.) What type of exits did you use? 5.) Can you think of anything that I'm missing? (Sometimes one gets so close to the trees that he doesn't see the forest, and sometimes one get so close that all he see is bark, and forgets there is a tree there.) 6.) What were your risk:reward parameters? What were your actual returns and risk? 7.) Would you recommend trading in the consolidation phase of the markets, or forego that and simply trade the trending phase only? 8.) If you recommend trading in the consolidation phase, what type of system do you recommend to do that? Trending phase trading appears to somewhat of a no-brainer, but I'll be damned if I can figure out how to get into those smaller (5-6 point) moves without getting whipsawed half to death most of the time. Thanks in advance for your (or for anyone else's, for that matter) help. DGBrothers
DGB. Please keep us up with how you are doing with this once (if) you start trading it. Ehler's work is something I have on the pad to have a look at once time permits, so I would be interested to know your experiences, and whether I should move it to the top of the to-do list.