## Building a Forex Portfolio With the Highest Possible Level of Robustness: Introducing Comitl

During the past 6 months I have embarked on a quest towards the development of what we might consider “very robust” trading strategies. This journey started because I noticed a general lack of highly inherently robust strategies within Asirikuy, something which prompted me to develop systems with very high standards regarding robustness. During the past two months these efforts have been concentrated on a new EA – called Comitl – which is a portfolio based solution that does not take robustness as a priority but as a characteristics of the utmost importance. Within this post I will talk to you about this new EA I have developed for Asirikuy and why it is different and much more robust than other Asirikuy strategies. I will talk about some of the characteristics of this EA, how it may be improved and what I expect from it in the future.

Any algorithmic trader will carry with him or her the inevitable burden of trading uncertainty. How will we know if your strategies will fail now or in 20, 30 or 50 years? How do we know if the targets we have considered realistic – through Monte Carlo simulations and careful analysis – will indeed materialize in reality? These questions are very important and constitute the basis of perhaps the most important issue within trading: survival. Developing systems with survival in mind is something absolutely important which needs to be given a high priority when building and testing strategies. The truth is that there is an inherent compromise between possibility of failure and expected profit and therefore a good trader needs to carefully gauge both factors. While new traders tend to focus on maximizing simulated profit – almost always ignoring issues pertaining to robustness – conservative traders look in the complete opposite way, sacrificing profits if it means an increase in robustness.

I wanted to develop the next Watukushay EA thinking as the most conservative of all algorithmic traders would think. I wanted to build a portfolio seeking the highest level of robustness with profitability being only an afterthought within the whole development process. With this in mind I decided to develop a portfolio for the daily time frame which would have some very interesting characteristics. The portfolio would use three systems, all systems would use the same settings on all pairs and all pairs would have twenty years of backtesting on the daily charts. No optimization, no hind sight, no parameter changes, no fitting at all. The twenty year period can be viewed as a pure out of sample evaluation as the strategy was never optimized either on one or many of the pairs used. Even when doing the initial design of the strategy – based purely on visual chart observation – it was done only for a few years within 2000-2010.  Also – in order to avoid portfolio selection bias – I decided to trade the 4 most liquid pairs in 1990 (which are the ones traditionally known as the 4 majors).

The portfolio EA – which includes the three strategies inside of it – uses two systems I published on the February and March issues of currency trader magazine, using Keltner channels and Bollinger Bands and a separate price based strategy using a time determined exit. As I have mentioned before the parameter selection for these systems was done one time through visual observation and it was never changed nor optimized through any trading period. This portfolio EA therefore has a level of robustness which almost no other system can match: no optimization, 20 years of out of sample testing, four pairs and the exact same settings on all of the chosen currency pairs. No portfolio selection bias, no optimization bias and therefore simply no risk at all of curve fitting.

I was happy with this achievement since it constitutes what many would consider a system that works at a very fundamental level. It works through several pairs, it works for long periods of time and it has no problem with issues related with curve fitting. This portfolio has – therefore – an incredible natural chance at future survival since it seems to exploit an inefficiency which is simply a part of at least highly liquid instruments. Another important point is that the number of total trades within the 20 year period is more than 2000 (thanks to the use of 3 systems) and therefore the results are indeed statistically significant and put the system far away from what could be considered profit from mere random chance.

Comitl – a word in Nahuatl (language of the Aztecs) – means “pot” because this portfolio mixes three different systems which achieve good results using the exact same parameters on several pairs. As the different ingredients mix within a pot to form a single soup, so do the systems of Comitl naturally mix to form a very long term and robust profitable system. Certainly Comitl does not achieve the level of profitability of many other Asirikuy systems – as it sacrifices a great deal of profitability for robustness (AMR to Max DD ratio of about 0.4)  – but it has the highest inherent chance of future survival amongst all Asirikuy strategies. It could be well considered a very long term investment vehicle and it might even work across a wide array of highly liquid instruments (as futures, bonds, etc), as each strategy has the characteristics of a multi-market successful system.

Another thing which arises is the possibility to build alternative Comitl setups which have a lower compromise with robustness. Certainly someone may decide to optimize Comitl on certain pairs or trade the system with an optimization for the past 11 years, a valid compromise someone can make between profitability and robustness. The achievement here however is that Comitl grows from the base of possible profitability without any compromising of robustness, it grows from the assumption that survival is the most important aspect of trading (above profitability). Therefore Comitl’s base settings are the strategy’s “zero compromise” parameters which can be improved towards the profitable side with a loss of its natural robustness. As I have said before it is up to each trader to decide how much they want to risk for an expected profit and Comitl is the first system which might allow a fine tunning of this from a base level where no compromise is inherently present.

### 17 Responses to “Building a Forex Portfolio With the Highest Possible Level of Robustness: Introducing Comitl”

1. Daniel-

Looks great and i’m looking forward to checking it out in demo and possibly live. Keep up the great work,

Chris

Hi Chris,

Thank you very much for your comment :o) I’m glad you like the idea behind Comitl!

Best Regards,

Daniel

2. Marc says:

I ran across your site the other day, and there is some good information here. Some I take issue with, but its mostly nitpicky things. The one thing that I’ve read that I completely disagree with is the notion that scalping (or intraday) automated systems don’t work.

I think the reason that you haven’t gotten it to work is because you trade Forex. Its about $45 in total trade costs (comm + slippage). There is no way you could make money intraday because the average 30 minute range of Forex is like$150. You need to be in all day just to get enough movement in your direction to overcome your losers. If you trade 6E/EC CME futures, your total cost would be about $15 (comm + slippage). That’s a lot easier to overcome, because now your win percentage doesn’t need to be nearly as high. A 5 pip loss ($10 / pip) goes from being ($95) to ($65), and a $100 win goes from$55 to $85. Being an automated trader, I’m sure you appreciate how big those swings are in terms profitability, and I’m pretty sure you could take more trades and make more money. Regardless of your strategy or if you disagree, you probably need to look at CME Currency futures to see if they are liquid enough for you. Why would you pay all that money in Forex commissions when you don’t have to? Its money straight into your pocket for opening a new account. • admin says: Hi Marc, Thank you for your comment :o) I fully agree with the things you’re saying and I personally speak about the feasibility of scalping in retail forex. Certainly profitable scalping becomes possible with low commission regulated exchanges (where there is no feed dependency and you have access to data with market depth) but on Forex it is practically impossible, at least at the retail level (something I think you agree with me on). Regarding why I don’t trade futures well it relates to several aspects not dealing with which market is better but with minimal capital requirements, account mangement flexibility, general population access, etc. Forex simply has a lower level entry which you don’t have in futures something which allows me to diversify amongst tons of different strategies without needing a gigantic amount of capital. Even if I traded micro futures I would be limited to a few strategies while on FX I am hugely diversified with the same capital. I will certainly write a post about this in the future as I have answered it quite a few times :o) Thanks again for your comment, Best Regards, Daniel • Marc says: I’d never trade at greater than 10:1 leverage, so I guess I never factored that in. I’m fairly conservative, so its about the same either way for me. Why are you diversifying strategies so abundantly? The more you diversify, the more you are just averaging yourself back to the norm. Isn’t that why you should just pick your best one or two and run them on the best securities for them? For example, Its like being an active stock trader and owning 50 stocks. Couldn’t you just buying the index in the direction you are leaning? Unless you have analyzed the market and concentrate the money in a few out performers, you won’t beat the market. You’re essentially doing tons of research to find the outliers. Why not just go with whichever outlier you deem best instead of averaging it down? I know this that tone doesn’t come across well in text. I’m actually curious about your thinking and in no way saying you’re wrong. I’ve wrestled with these same issues and come to different conclusions. • admin says: Hi Marc, Thank you for your comment :o) I do agree with some of the things you have said, however the problem in my view tends to deal with the possibilities of future survival. If I stick to only trade systems which have been great then I am putting myself at greater peril regarding their potential future failure. There is always a possibility of system failure and therefore diversification tends to play a role in diminishing this possibility. However I do not believe that this in anyway reduces the possibility to “beat the market” on the contrary our research shows that the trading of different strategies in portfolios increases overall performance while also diversifying your trading. The more inefficiencies you tackle the more robust your portfolio becomes as a whole, this seems to be – at least – our conclusion from Forex trading experience. Certainly I do not think that when doing this you tend towards any “index” -as if you purchased many stocks- but you tend towards the market’s inherent compounded yearly profit to maximum draw down “limit”. In the end this is a game of measuring the possibility of future failure against expected profits. Sure, you could make an extremely good portfolio by taking the systems with the best historical performance on the best securities but then you run the risk of having a higher chance of future failure. In my experience both things need to be compensated and you need to compromise with a level of selection, robustness and expected profit which satisfies your risk appetite and trading personality. I hope this answers your questions :o) Thank you very much again for your comment, Best regards, Daniel • Marc says: Do you have the research on the porfolios or is it confidential? I’m assuming its probably a bell curve with a peak of the optimal amount of strategies to run. I’d be interested in seeing what 1 standard deviation form the mean turns out to be. Is the optimal number 3 or 20. It’s be interesting to know. I’m not sure I follow this. “Certainly I do not think that when doing this you tend towards any “index” -as if you purchased many stocks- but you tend towards the market’s inherent compounded yearly profit to maximum draw down “limit”. ” I think this is indexing. Its just indexing to Forex. If you are tending towards the markets inherent yearly profit and maximum draw down limit then that is approaching matching market performance. In essence, couldn’t you just buy a basket of Forex securities with no strategy and do that same thing? • admin says: Hi Marc, Thank you for your reply :o) The information we have regarding portfolios is available within the Asirikuy community so feel free to join if you want to access the results of my research and use our strategies. Regarding the Index, buy-and-hold strategies don’t work in Forex because the market has no fundamental bias (like stocks) there is no fundamental reason to believe that the market will go a certain way and therefore there is no “no strategy trading” to which you can compare your systems to. For example you could not simply “buy a basket” because there is no fundamentally known future (why would you buy USD over EUR, etc. In stocks there is a fundamental upside bias, in Forex there is no reason why a currency MUST get valued over another). The concept of “market performance” cannot be applied to Forex . Also the number of systems to form portfolios that achieve the best performance is not easily known since you can reach the same AMR to max draw down historical results by using either 4 or 10 systems – for example – depending on how well the systems match between each other (how well they hedge each others draw down periods). Certainly you also need to consider that the 4 and 10 system portfolios in the above example might have different characteristics which might make you choose either one despite the fact that they have a similar performance regarding profit to draw down ratios. For example the 10 system portfolio might have smaller draw down periods while the 4 system portfolio might have higher probabilities of having profitable months. I believe that the problem here is complex and definitely having as many strategies as you can with intelligent combinations that favor robustness and expected profits in different measure is the key to long term success with profits above the market’s performance (measured as the S&P 500 buy and hold return). To sum it up I believe that the development of strategies to tackle various inefficiencies and to address different levels of robustness is vital and the building of portfolios has already demonstrated that this is a much better strategy than using simply the “best strategies” based on historical performance. Having many systems ensures a high degree of robustness, a high probability to survive under future market conditions. I hope this helps :o) Best regards, Daniel 3. McDuck says: I guess the ultimate robust system would be to add to the portfolio a stock index, a bond index, and a gold index (and maybe a real state index). In 1-day bars there is probably reliable data from 20 years ago and more. • admin says: Hi McDuck, Thank you for your post :o) You’re very right about the availability of very long term data on indexes and one of the things I want to do is to build both Coatl and Comitl portfolios including these instruments. Right now I only have data for gold and silver but once I get my hands on long term EOD data for other instruments I’ll be on my way to what you mentioned :o) Thanks again for your post, Best Regards, Daniel 4. Marc says: One other thing that I disagree with. Strategy development is curve fitting. Just because you did it the hard way (visually) doesn’t make it any less curve fitting. So long as you iterated through a process with increasingly good results based on observations (visually or mechanically), its curve fitting. You’ve got a big out of sample, so it doesn’t really matter. Its just impossible to not curve fit a strategy. If it wasn’t fit, it wouldn’t have a positive expectation based on the history. • admin says: Hi Marc, Thank you for your comment :o) I completely agree with you in that there is a certain degree of hindsight in the design of all strategies as you have in fact “known the past” when you build any system. However my point was that the strategy is void of “curve fitting” introduced by optimization procedures which often make the problem much bigger. Certainly – as you said – there is an inherent fitting degree to all algorithmic strategies, regardless of how much optimization they are subsequently subjected to. Thanks again for your comment, Best regards, Daniel • Franco says: Hey Marc, I agree curve fitting is a problem, but the curve fit bias becomes less and less the more trades are executed in the specified test period. For instance if you ran a 10 year backtest, and 1400 trades were executed with a profit factor of at least 1.2, then the chances of that system being curve fitted is almost zero. On the other hand if your system only had 80 trades then the chances become much more for a cruve fitted system. That is why in my opinion it is important to develop systems that trade often, to remove most of the curve fitting bias. This is all of course just my opinion, I don’t have live testing proof for this but it is a statistical sound principle. 5. Chris says: Hi guys- Marc – I appreciate your commentary of Daniel’s work and the perspective you bring to the table. What i’m trying to achieve from Daniel’s works is to gain a competitive return on my assets say versus a passive investment in the S&P 500 index. Also, I want the returns to have minimal correlation to the S&P 500 returns. Another point is that aside from the analysis on weekends etc, this is a passive investment. In other words, I can go about my life and full time job without studying charts and trying to scalp 30 minute moves. If you are able to acheive consistently high returns with manual scalping then congratuations and by all means continue. My goal on the other hand is to diversify my investments to Forex and employ systems that are well-researched and statistically sound. These systems may not beat the best real-time, hands-on traders, but I expect they will be competitive versus stocks, bonds, etc. I might also make the case that full-time trading is a highly competitive business and success does not come easy. You can follow the evolution of my trading using automated strategies on my blog shown below. Regards, • Marc says: I’m an automated trader, and I never do any discretionary trading. I have a real job as well, so I can’t. I also never have positions open over night. I can’t sleep well knowing that I could lose power or something. My whole issue was that its possible to scalp and it doesn’t take intense chart study once you have a working system. For instance, look at high frequency traders. They are making money hand over fist, and its doubtful they are doing any chart study over the weekend. This flies in the face of the logic that scalping strategies don’t work. The reason why HFT works for them and not retail traders is low fees (no data feed fees, no slippage due to co-location, and minimal trade costs). Under that scenario, a lot of people would have working scalping strategies, and its the reason why there is a lot of complaining on the Street about it. Its an unfair advantage. To me, there is no difference between a swing strategy and a intraday strategy in terms of research and maintenance. Once its on, its on and you just need to monitor the statistics. If you reduce your costs, you can have a lower winning percentage, trade more, and increase your overall next profit because now your acceptable gain is within the expected volatility range of the intraday market. The lower the costs and higher the volatility, the smaller the time frame that you can exploit. I also think that your risk/reward is off. A Forex Contract is akin to having 100K at your disposal with$2,000. Therefore, your expected return should be on the 100K. So if the market goes up 13%, you should make $13,000 (100K *1.13) not$2,260 (2K*1.13), correct? If you only expect \$2,260, your expected reward for the use of leverage is way too low. You’ve skewed the risk/reward the wrong way. Did I misunderstand what you said?

6. erick says:

Always good to minimize the possibility of account wipe-out, for obvious reasons. But I wonder if “trading to not lose money” misses the point of trading, which is to make money.

I understand that a wipe-out is always possible and always a concern because the possibility is fundamentally unknown and scary, even knowing the statistical AAR/DD ratio.

My idea is that it might be better to set a return target of, for example, 30%/year and then minimize risk as much as possible down to this return constraint.

That is, first prioritize a return target, then work on the risk management.