Reverse Engineering a Profitable System (REPS)

No I don't think you're wrong about backtesting at all. I just think we have to define what we're backtesting.

And you are correct about this scenario. Almost perfect. Cherry-picked. But why is that a bad thing. Why do we always come up with a trading system and then hope it will work well. That's my point in "reverse" engineering. Design something that would have worked well in one of these "near perfect scenarios" and then use that moving forward.

As far as entering in trends at the early stages, we've established the break of the DC on the weekly would signify a new trend. Through a DC with a setting of 20 on a weekly chart and see how well that would signify the start of a tradeable trend. Granted, we'll start many that will either hit stoploss or perhaps get taken out at BE, but that's okay. That's why we're keeping our initial risk low so that we can be around for the long haul and be there when the super trends hit. There not as rare as you might think.

Okay so our next entry appears on 9/17/14 as prices pushes down the DC at 1.2857



As we move to the H4 we see our top of the DC s/l at 1.2995 or 138 pips. Little bit larger stop but...that's where it is.



If we move our previous entries to that s/l as well we see we have floating profit on our positions as follows:
Initial entry: 512 pips at .08 lots for $409.60
2nd position: 329 pips at .05 lots for $164.50
3rd position: 159 pips at .20 lots for $318.00

Total running profit is $892.10. We want to risk only 50% or $446.05
Divide that into our desired s/l and we get a position size of .32 lots.
1 week later we get a nice move down and we get to move our stop to BE on this most recent entry.

Now we have a total of .65 lots running at $6.50 per pip. Again all risk free with a guaranteed profit of $446.05

It's also important to note that now our running stop loss on our previous positions is at the upper level of the Daily DC. So now our series group will not get closed unless we get a series ending hit of the upper daily DC.

You may also note on the daily that price action in the middle of October is going to take out this most recent position at BE. That's okay, we'll still be in the series, but we will loose one of our positions. But we'll get to that later.

So for our last position we will go into detail on is to be found at the 1.26478 level on the 30th of September.



Our s/l according to the H4 DC will be at 1.2784 for 137 pips
Remember we don't want to bring the trailing stop on our positions inside the daily DC now so for this trade the stops on positions 1-4 will stay where they are.
So for this trade the floating profit we have to deal with is the same as before. $446.05. Since our desired stop happens to be almost exactly the same as trade 4, the position size on this trade will also be at .32 pips.



After a small retracement price makes a big move down on October 3 and we move our stop to BE.
We're now carrying .97 lots with a value of $9.70 per pip. Again at no risk and a guaranteed $446.05 of profit from the series.

Again we notice the price action is getting ready to take out trades 4 & 5 at BE. So be it. We won't have any more entries until at least October 31st when the bottom of the daily DC is again taken out.

I hope you get the idea thus far.
As I haven't heard any other options to change anything I'll figure out the rest of the entries etc. (as I have time) for the remainder of this trend and post the results in full.

Happy pipping until then.

Hi Spyderman

Just wanted to let you know I'm really enjoying reading along with your project / experiment

Please take this as far as you can as personally I always end up learning something new from those who are trying
to look at things from a slightly different perspective.

Adding on to trades has always been a sticking point for me but the 50% method is new for me and I like it.

cheers

Sunny

Usually these big swings are on diverging monitary policy... For example Aud entering a rate cut cycle... or Usa planning to raise rates, while ECB starts printing. Or Jpy starts printing endlessly. That's the fundamental driver that pushes these swings so far so fast.

Okay, so I ran the possible trades through to the DC exit.
Some got closed out at BE along the way.

Here's a screen shot. Prepare to be skeptical



We ended up with 7 entries. When closed at s/l they shook out as follows:

1st entry at 1.3505 for .08 lots. Total of 2,470 pips. Profit - $1,976
2nd entry at 1.3333 for .05 lots. Total of 2,300 pips. Profit - $1,150
3rd entry at 1.3152 for .20 lots. Total of 2,120 pips. Profit - $4,240
4th entry at 1.2247 for .36 lots. Total of 1,220 pips. Profit - $4,392
5th entry at 1.2066 for .82 lots. Total of 1,033 pips. Profit - $8,470
6th entry at 1.1750 for 1.04 lots. Total of 720 pips. Profit - $7,488
7th entry at 1.1270 for 6.23 lots. Total of 240 pips. Profit - $14,952

Total profit - $42,668 with an initial risk to capital of $100.

Obviously we were fortunate to not hit any s/l's along the way and got all our trades to BE.

Let the flaming begin...

The deal is to keep things above water until this kind of run ... So no mater an early exit ... it will happen most of the time ... The nice run will make this worth.

Have you seen what Copernicus is doing with respect to SDC channels for longer term trends?

Might be worth a look - at least from the point of view of staying in the trend.

http://www.forexfactory.com/showthread.php?t=514163

Perhaps this is not the right time to ask, but I'll ask anyway

How will you decide which pairs to watch?

I know it must be taken with a grain of salt but could it be useful to
scan through some of the larger banks 2015 2016 forecast prices and see which
pairs might potentially have the largest moves forecasted.

just a thought

A tough question. The long term trend opportunities for the last 15 years or so lie in XAGUSD, XAUUSD, EURUSD, GBPAUD, DJI30, S&P500, GER30, BRENT, BUND et al. They appear to be commodity, indices and treasury note heavy but we do have some eligible curency pairs such as EURUSD, USDSEK, USDSGD.

AUD pairs, CAD pairs and NZD pairs in general (excluding GBP/USD related pairs) have been woeful and signficantly impede portfolio performance....however NZD pairs are showing some recent trending opportunities.

Granted this is from the benefit of hinsight but if you wish to be instrument agnostic then I would suggest you test various instruments using a roll forward back-testing approach. For example, back test years 2000 to 2003 for a particular instrument and if it stacks up as a trender during this period then run a test blind for the 2004 year or until the trend you are on ends. Then rinse and repeat using say 2001 to 2004 to test 2005 etc. etc. This way you are avoiding the propensity to curve fit and more fairly assessing your trend following portfolio approach.

The key requirement for long term trend following approaches is risk management during non-trending periods. You never know when a trend sets off so you need to be ready to trade when your system tells you to....and drawdowns are a necessary evil....so risk management is crucial. That's why I always side on risking a very small % of risk capital per trade. There were extended non-trending periods between 2000 and 2010. The last 5 years and in particular the last 2 years to current times have been tremendous for trend following strategies. It is a bit misleading to backtest from 2010 on as this biases on the plus side trading performance of trend following strategies.

Sounds good to me, but i'm a little bit skeptical not because of the theoretical earnings but because of the near perfect scenario this system needs.

Anyway, this is a perfect candidate for an ea. All rules are objective and if you want to apply this to multiple ccy pairs you need it.
Also, how to manage the MM without an ea. Would be very difficult.

I have two other questions:

1) the first one but seems obvious, just for confirmation: so we enter only when the closing price is below the donchian right?
In this case we need a fix, because a lot of times what we have is a trend in which the new highs just move up the donchian but candles don't close well above/below it. So each new day we have a higher dc but price can't close above it (well, the dc of the day before, being impossible that the price closes above the current day high).

2) If we have 10 ccy pairs and we want to enter in 5, how to use the unrealized profit? They will belong to all of them, should we count just the ones coming from the current currency pair we want to jump into?

Thanks man, your effort is admirable.

True story. Absolutely vital to keep the risk low at the initial stage so we're still in the game when things get interesting.

Throw a DC on some weekly charts and see what they reveal. Here's the 1st 2 I pulled up.




Do you see tradeable trends there. Granted the E/U we used is an extreme example, but again that was to define a strategy. There are much shorter trends that will still yield profits.

I'm going to start testing the waters this week. It will of course take a while to get this rolling due to the long term nature of the system. Plus this is a notoriously slow time of year. But we have to start somewhere.

I have a live account that I'll be attaching to the thread as an explorer.

I'm going to request FF to put this thread back into Trading Systems or maybe Journal.

Hey Lemming. Thanks for the questions.

1.) I use the break of the DC channel and maybe add a couple pips. We're looking for momentum breaking a previous high/low, so not concerned about the candle close.

2.) Yes each pair series stands on it's own. So you might have a series in trend that's trading some larger positions while you are starting testing another pair with your 1% or whatever you decide on.

Copernicus,

Thanks for you comments.
I agree whole heartedly on the MM aspect. Have to keep the initial risk low. 1% may be too high. Ultimately, I'd like to be in the .25% area on initial series trades.

Unfortunately I'm starting with a smaller account of $1k equity. So with a minimum position size of .01 lots I'll be risking a bit more depending on my s/l sizes.
If things go positive I'll be adjusting that down.

Hi spyderman. Nice thread. :-)

We are on similar missions.

The shorter term Donchian exit within a longer term Donchian entry filter is a robust method that certainly keeps you out of the rough and ensures you capture any trend of significance.

If it is any help, these are the findings I have come across over pretty extensive testing and they may be useful for your strategy.

1. There is a direct relationship between increasing noise and the timeframe use, being that preferable timeframes for trend following now need to be longer (as opposed to a shorter duration). The sweet spot currently on D1 with most traditional trend following methods appears to be approximately 200. The reason for this requirement is one of avoiding noise through attempting to identify legitimate price action bias. This is also evident when looking at various timeframes for trends and seeing a direct relationship between better trends and longer timeframes. The longer timeframe serves to smooth the noise. This therefore has implications regarding the settings of trend filters that seek to eliminate noise and capture volatility such as Donchian entries, moving averages etc. Increased use of trade algorithms, platform technologies and globalisation is increasing randomness at the short term timeframe and to avoid the churn, you need to work outside this noise.

2. Trade entries are not as important as methods to keep you out of the rough. It is easy to lose track of this principle with trend trading and think that the entry is important. The Donchian trade entry filter is powerful not because of its trade entry definition but because it keeps you out of sideways congestion which hampers any trend following strategy. For example, if you have a filter that can eliminate trading the sideways congestion, then other trade entry techniques such as trade retracement entry can be particularly powerful as it allows you to leverage position sizing while keeping within trade risk % requirements. In fact I would suggest that any trade entry into a trend is as good as any other (provided of course that you can keep out of the rough).

3. Anyone in hindsight can identify trends that they 'should' have traded, but in working on the right hand side of the chart, while respecting the principle of letting profits run and cutting losses short, it is a difficult exercise in achieving this balance and staying on a trend for the long term. Random price movement (noise) is the greatest enemy when you are also attempting to minimise risk. You need to develop trade management methods that allow price movement to breathe but in turn, expect to sacrifice performance in achieving this aim. Overtrading and intervention kills trend following.

4. Use of moves to breakeven etc. as a way to manage trade risk as opposed to using a primary exit mechanism such as a break of the shorter period Donchian or a MA crossover etc. are a two edged sword. Yes, they manage risk.......but they also significantly compromise profit performance. It is a tough decision deciding what to do and I have used long range testing as a basis to determine my stance on this. I have found better results in terms of lower drawdowns and improved returns from interfering less with trades. Testing is demonstrating to me that the primary exit mechanism should be the only tampering you do with trade management. For example, use the short term Donchian exit as a basis for stop setting and position sizing. Locking in moves to breakeven while in principle sounding good often disrupts potential longer term profitable trade entries. Given random jitters that are built into price movement, letting price breathe is crucial for trend following.
Interestingly there appears to be a direct relationship between performance and trade tampering. The less you tamper, generally the better. This has deeper roots in the principles of overtrading which are a trend traders worst friend.

5. Adding to trades in trend can also be a two edged sword for performance. Where there appears to be merit is when you add to trades early in a trends development. When the trend ends, those recent trade additions inevitably get clobbered and sacrifice total performance. You might want to consider maximising the number of trade entries you employ for any given trend. The principle of only adding positions only when the prior trades are profitable if closed out is a good one but consider limiting entries to a maximum number using this principle. While there will be a few instances where some excellent long term trends allow you to pyramid extensively, the impact of continuous trade addition on trends means that you raise the benchmark on the required trend conditions you need to effectively harness this principle. It will work in your favour with some beautiful trends.....but they are few and far between. The profit potential on most other trends will be significantly reduced with continuous trade additions.

....anyway, best of luck spyderman. I don't think you are going to need it however as your logic and intuition in applying this to trend following are shining through. :-)

Copernicus,

Many thanks for the insightful remarks. I'll definitely digest this info. You've obviously spent a lot of time on this. Keep stopping by and checking on our progress.

I'm in and will continue to participate if I can add anything helpful. There are some excellent trend followers in this forum. I have often wondered what we could do if we all closley collaborated and woked collectively together. It is a hard enough game as it is......... :-)

Hello Spidey,

IMHO that is a mistake. You are aware that random data can produce trends just like you have cherry picked? Without any stats to design a system around, how can you have faith to last through the randomness? You would do better if you started with stats and then explored systems around those stats - then you are moving outside of the randomness and shifting probability in your favour. This is not reverse engineering.

With kind regards,
MK

MK,
Appreciate your thoughts. Have to say I disagree however. You speak of randomness producing trends as if that would somehow discount the opportunity. There's much debate here on FF and elsewhere as to how random the market is or isn't. As I stated earlier, I just don't care. It is immaterial how/why long tail market movement happens to me, the point is that it can be traded. Those trends happened, random or not, and they will happen again.
The "stats" that are important to me is what the price did in fact do and, at least in my mind, represents what it will do at some point on some pair. It is around those stats, market movement, that I explored this system.
To me, we are reverse engineering in this respect. Typically I see folks design a system and then backtest or forward test to see how it does. We are taking an end product, a sustained market movement in a given direction (trend) and then designing the system from that. Then I can have a measure of confidence that, when another sustained market movement comes, my system can and will trade it with a high profitability.

Cheers