In this article we present a simple trading system that we developed based on the concepts outlined in The 10 Power Principles of Successful Trading Systems.

Step 1: Selecting a market and timeframe​


One of the most popular markets these days is the e-mini S&P, and that's not without a reason: It's a 500 company index. One of the largest in the world and that means you have excellent and consistent liquidity, superb volatility, tremendous leverage and no uptick rule. It's a truly bi-directional market that shorts just as easily and safely as going long. It's a fully electronic market, offering all the advantages of electronic contracts.

We decide to trade the market intraday, i.e. we will enter and exit a trade on the same day, because we do not want to expose our position to the risk of holding it overnight.

Step 2: Define entry rules​


In my opinion swing trading is actually one of the best trading styles for the beginning trader to get his or her feet wet. That's why we decided to use a swing trading approach in this example.

Many traders are familiar with the concept of "Bollinger Bands": Bollinger Bands consist of a centerline and two price channels, one above the centerline and one below. The important thing to know about Bollinger Bands is that they contain up to 95% of the closing prices, depending on the settings.

pic_01.jpg


In the chart above you see the red centerline and the blue price channels. There are only 2 days in the beginning of November when prices close outside the Bollinger Bands.

We are using this knowledge to create a very simple entry rule:
  • Sell when prices move above the Bollinger Bands and
  • Buy when prices move below the Bollinger Bands.

The idea is that prices will move back into the Bollinger Bands by the end of the day.

Step 3: Define exit rules​


Let's start with a very simple exit rule:
  • Exit the trade at the close of the same day.

Below is the equity curve of the last 2 years. The first results are encouraging.

pic_02.jpg


Step 4: Evaluate your system​


The Net Profit of this simple trading system is $13,525.
The Average Profit per trade is $149. Even if we deduct $20 for commissions and slippage we still have a net profit of $129 per trade.
The Profit Factor is 2.20.
The Winning Percentage is 66% and the Maximum Drawdown at the end of the day is only $2,775, though we have to suffer an Intraday Drawdown of $5,250.

The next step is to test the robustness of the system. Therefore we will vary the parameters we are using for the Bollinger Bands to make sure that we haven't curve-fitted the system. If the system produces similar results when we vary the original parameters by 15%, we have a quite robust trading system.

Originally we tested the system with a setting of 34 for the Moving Average and 2.5 for the Standard Deviation. The table below shows the results of the system when using a Moving Average between 29 and 39:

pic_03.jpg


As you can see, none of these figures change dramatically when varying the parameters.
In the next step we run the system on different markets to make sure that we haven't optimized the system for a single market.

We test the system on 5 different markets:

pic_04.jpg


The net profit, average trade and max drawdown are substantially different, but the Profit Factor seems to be quite stable. The reason for this distorted picture is the different value of these five markets. In the next table we look at the Average Profit and the Max Drawdown as a percentage of the net profit:

pic_05.jpg


Now we see a different picture: Only the Max Drawdown differs quite a bit depending on the market, but the remaining figures are rather stable.

It seems that we developed a robust trading system that will perform well in real market conditions and on several markets.

Step 5: Improving your system​


We try to improve our system by adding a stop loss:

pic_06.jpg


Notice that the system performs best without any stops.

Another interesting test is to increase the duration of the trade: The original rules said that we exit the trade at the end of the day. The following table shows the results when we add x days:

pic_07.jpg


If we exit on at the end of the 2nd day after entering the market, we increase the net profit and decrease the Max Drawdown. That's the kind of improvement we are looking for.

As a last step we test these settings on the five markets again to ensure that we haven't curve-fit the parameters to only one market:

Original settings (exit on the same day):

pic_08.jpg


Modified settings (exit on the close of the 2nd day after entering):

pic_09.jpg


We can see a dramatic improvement in the other markets, too.

Conclusion​


We started with a very simple idea and defined two easy entry rules. Applying the simplest of all stops (exiting at the end of the day) we received a system with a nice performance. We tested the system with several parameters and on several markets to make sure that we haven't curve-fitted the system to a certain parameter set or market. Then we tried to improve the system: While applying a stop loss did not increase the system's performance, the increase of the time in the trade was very successful: We increase the net profit by 30% while decreasing the max drawdown by 17%. Applying these new rules to the other markets we noticed a dramatic increase of all figures in all markets.
 
Last edited by a moderator:
Congratulations Mr. Heitkoetter! You have given birth to a beautiful curve-fit system.
 
Markus,

May be I am a bit slow - as usual: But I don`t understand the "improvement" on the s&p numbers :

You start with a net profit of 13,525 and end after optimisation at 8,100 ?

Furthermore your rule no 5 recommends five trades per week: Would you prefer to start one swing trade per day - or diversify over other markets?

Thanks

Hittfeld
 
Last edited:
Spam Man said:
Congratulations Mr. Heitkoetter! You have given birth to a beautiful curve-fit system.

Hello Spam,

I may have missed the point here (probably have :eek: ) but the system uses a single parameter (or 2 if you include the stdev), at the same setting, over a number of markets.

Why do you consider it to be curve fitted?

Cheers,
UTB
 
"SIMPLE" - that is the key!
keep it simple and it works
Get hundreds of indcators - u get confused
 
Hi UTB,

The curvefit accusation is quite interesting. Although I think the principle is quite valid for index trading (and could even be improved by scaling in as it moves away from the entry point) I also think that you have more parameters than you realize:

- moving average length
- std deviation for the bands
- stop loss size
- holding period

I don't think its curve fitted but I would test it on some different period samples as well as different equity markets because the equity markets are highly correlated (in volatility as well) and you may find that they dont give a good test. Try the HSI as well. Regarding the stop loss I have rarely found the SL to improve a system but it often reduces the variability of the trade outcomes in a useful way.
 
the blades said:
Why do you consider it to be curve fitted?

Run enough backtests and you'll come up with plenty of systems, both simple and complicated, that work well in the past but fall apart during forward-testing because they don't have an edge. That's not to say that someone trading this system exactly as is would definitely lose money - but if they didn't it would be down to good fortune.

If a backtest was run over time periods other than that shown in the article, my guess is the performance would be significantly different, and probably worse. The author is welcome to prove me wrong.

To be honest I wouldn't know where to start trying to find an edge based on Bollinger Bands. It would require the market to display a kind of stationarity which doesn't exist.

Incidentally, aren't Bollinger Bands drawn after the close of the bar? If so, you wouldn't be able to trade this anyway. Perhaps there's some kind of offset that I don't know about.
 
Hittfeld said:
Markus,

May be I am a bit slow - as usual: But I don`t understand the "improvement" on the s&p numbers :

You start with a net profit of 13,525 and end after optimisation at 8,100 ?

Furthermore your rule no 5 recommends five trades per week: Would you prefer to start one swing trade per day - or diversify over other markets?

Thanks

Hittfeld
Hitfeld,

When "optimizing" a system I don't try to "maximize profits"; I am trying to improve a couple of key figures. In this particular example the net profit decreases, but the average profit per trade increases and the max drawdown decreases. Overall the system becomes more robust.

IMO you can supercharge a robust trading system with proper money management techniques, even if it is only slightly profitable. More important is a smooth equity curve.

Regarding your second question:
Yes, the system does not meet the requirements of rule no 5. Despite the comments of SpamMan I don't think that it is easy to find a system that is profitable, robust and meets all the rules outlined in the 10 Power Principles. But being very selective when developing or purchasing a system will increase the probability of success. :)

Markus
 
Are we skeptical about the system because of its components? ... back test process? ... both?

On paper, the simple system described would appear to be easily fooled by trending markets, as is true for many oscillator-based systems. It's not uncommon for prices to be outside their Bollinger Bands for several days. The initial results of the sample system, though, show promise I did not expect.

I wouldn't trade this system before running it through more tests. Nor would I conclude it is "curve-fit" before determining that it is.
 
Spam Man said:
If a backtest was run over time periods other than that shown in the article, my guess is the performance would be significantly different, and probably worse. The author is welcome to prove me wrong.
SpamMan,

As I wrote in the article: I don't believe that there are any good "hybrid" systems that work in trending markets as well as in direction-less markets. IMO you use either a trend-following method that does well in trending markets and don't lose a lot in direction-less markets, or you use a trend-fading approach.

The presented Bollinger Band System is a trend fading method and worked well in the past 2 years because the indices weren't trending as much as they did prior to 2003. And here's why I think that the indices won't start trending again any time soon (but that's just another opinion, right? ;)):

http://www.rockwelltrading.com/rockwell-trading-articles/backtest_right_way.htm

To be honest I wouldn't know where to start trying to find an edge based on Bollinger Bands. It would require the market to display a kind of stationarity which doesn't exist.

Yes, the results are obviously different if you apply a trend-fading system on a trending market. That's why we still need traders with brains: You have to determine whether a market is rather trending (e.g. currencies) or rather direction-less (e.g. indices).

IMO this can't be done by computers (yet) and needs human judgement.

Spam Man said:
Incidentally, aren't Bollinger Bands drawn after the close of the bar? If so, you wouldn't be able to trade this anyway. Perhaps there's some kind of offset that I don't know about.
Most software packages draw Bollinger Bands "on the fly", i.e. they use the current trading price for calculating the Bands. As the close will be within 1-2 ticks of the current trading price the previous "estimate" is close enough for a trading decision.

Markus
 
Markus_H said:
As I wrote in the article: I don't believe that there are any good "hybrid" systems that work in trending markets as well as in direction-less markets. IMO you use either a trend-following method that does well in trending markets and don't lose a lot in direction-less markets, or you use a trend-fading approach.

That's a sensible comment - though I can't see it anywhere in your article.

Markus_H said:
Most software packages draw Bollinger Bands "on the fly", i.e. they use the current trading price for calculating the Bands. As the close will be within 1-2 ticks of the current trading price the previous "estimate" is close enough for a trading decision.

Let's not beat around the bush - this point alone invalidates your backtest. The price at which your backtest shows you entering a trade will almost always be different to that at which you would have entered if watching the bands being drawn in real time.
 
I think people are missing the point of this article. To me, it is the process of building a system, not that this particular system is any good.

Peter Schumacher
 
Maybe it's the late hour or that last beer but I don't understand the entry rules.

Buy when prices move above the BB
Sell when prices move below the BB

For the displayed time period (3 months), that would give us only 2 trades? What am I missing?
The rest of the time , prices move between the bands so no trades there?

Heasymo
 
Markus
I appreciate the effort but this simple trading system is for relative newbies to the trading game. And you're teaching them not to use stops?
If you backtest this further to 2001 you will find that you would have lost $5900 in 2 trades in 12 days on ES ie nearly all of the profit from this example gone in 2 weeks.
Capital preservation MUST be primary - profits are secondary IMHO
 
rdstagg said:
Markus
I appreciate the effort but this simple trading system is for relative newbies to the trading game. And you're teaching them not to use stops?
rdstagg,

Thanks for your comment. No, I am not "teaching" not to use any stops; my intention is to show and comment some results of developing and backtesting a trading system. After presenting the table I wrote "Notice that the system performs best without any stops." Just an observation. :)

Each trader has his own risk-profile, and it's up to the individual trader to decide how to use the presented results. When conducting webinars at the CME I frequently choose settings with a lower overall profit, but also lower risk.

rdstagg said:
If you backtest this further to 2001 you will find that you would have lost $5900 in 2 trades in 12 days on ES ie nearly all of the profit from this example gone in 2 weeks.
Capital preservation MUST be primary - profits are secondary IMHO
In 2001 the indices were rather trending, that's why you would have lost money when using a trend-fading approach. As I mentioned in a previous post: It is important to know the overall characteristic of a market before trading a system: In a trending market use trend-following systems, and in direction-less markets use trend-fading systems.

In 2003 the markets changed from trending markets to direction-less markets. That's why you will see worse results in 2002 and earlier.

Markus
 
Markus,
Thanks, I enjoyed the article. More ideas into the pot IMO. I have it running on TS and personally was not disturbed by how it backtested. I have a portfolio that just tracks these public domain systems. It will be interesting to see how yours compares over the coming months.

I posted 2 equity curves back to start 1998. First is uncompounded second is applying fixed ratio MM with a $10k delta. The one obvious drawdown just re-enforces the necessity to diversify into many systems/markets and keep risk in any single strat limited.

Have included $2.50/side cost and $12.50/side slip.
 

Attachments

  • simple system c10kd.PNG
    simple system c10kd.PNG
    14.7 KB · Views: 457
  • simple system uc.PNG
    simple system uc.PNG
    13.7 KB · Views: 381
Markus

And therein lies the conundrum - can you tell me if tomorrow will be a trending era or a non trending era?
I would pay you handsomely if you could!!

My message was not meant to have a go at you - lord its difficult to post anything without someone disagreeing - merely to point out that what looks a great system (and it does have potential) to a newbie shouldn't be blindly followed.
And we all know that whatever the disclaimers some out there will adopt this system as is. Its human nature.

On the positive side I was very surprised at the win/loss ratio - so if you added this to a trending strategy portfolio it could be interesting in smoothing out the equity curve and even reducing drawdowns.
Perhaps thats the subject for another article?


Markus_H said:
rdstagg,

Thanks for your comment. No, I am not "teaching" not to use any stops; my intention is to show and comment some results of developing and backtesting a trading system. After presenting the table I wrote "Notice that the system performs best without any stops." Just an observation. :)

Each trader has his own risk-profile, and it's up to the individual trader to decide how to use the presented results. When conducting webinars at the CME I frequently choose settings with a lower overall profit, but also lower risk.


In 2001 the indices were rather trending, that's why you would have lost money when using a trend-fading approach. As I mentioned in a previous post: It is important to know the overall characteristic of a market before trading a system: In a trending market use trend-following systems, and in direction-less markets use trend-fading systems.

In 2003 the markets changed from trending markets to direction-less markets. That's why you will see worse results in 2002 and earlier.

Markus
 
Top