Meta-Trader - Coatl Review - Part 1

Welcome back Meta-Traders.

In this review, we head-on tackle Coatl which is a word in Nahuatl which means snake. The snake imagery will make sense as you get to understand the system. Be advised that there is a lot to know and understand here and it will take more than one review to cover this system and associated portfolios.

The image on the left is depiction of Quetzacoatl which is a flying reptile deity of the Aztecs. Looking at some of the images of Quetzacoatl, I have to think the Aztec were using some pretty powerful mind-altering drugs!

Anyway this system is Asirikuy’s first-ever system derived using genetic programming technology. While I am a computer science major, I had a bit of a misconception that when using genetic technology –that you could start by telling the system almost nothing, and it would somehow come up with a profitable system through the magic of computer science.

It doesn’t work that way, of course, and you have to give the system a starting point as well as a set of rule boundaries within which the genetic algorithm will search for a profitable system. In the end, Genetic algorithms are just search or optimization technique where the computer searches a huge set of possible solutions and comes up with a winner among many, many alternatives considered.

Before we get into the specifics, let’s set some broad boundaries:

- Coatl is developed and tested against the daily charts.

This allows optimizations to be performed against non-major pairs such as EUR/JPY, EUR/CAD and opens up the field of trading to many more pairs and longer histories. It does run on H4 and H1 charts as well, but on a much smaller number of pairs.

- Coatl is developed and tested using Control Point Simulations

I did some research on the difference between Control Point and Every Tick simulations. In a nutshell, it comes down to the number of times the Start() function is called during back-testing of your experts. Every-tick simulations try to reproduce intra-bar results whereas Control point simulations make fewer calls per historical bar. For daily charts used here, the results for Control Point simulations exactly match the results of every tick simulations, except they run many times faster. This allows optimizations to run many, many times faster which allows for optimization of a MUCH larger space than could be performed using every tick simulations.

- Coatl does not optimize moving average length and other indicator parameter values.

Instead, Coatl does very broad and coarse optimization of trading methods when applied to a particular pair. And so far the results have been spectacular for the EUR portfolio as you can over on the right-hand side of this blog page.

The actual genetic programming comes from Meta-Trader itself – simply check the ‘Genetic Algorithm’ check box on the “Expert Properties” screen inside the Strategy Tester window. When this box is checked, Meta-Trader uses an embedded genetic algorithm to reduce the size of the search space.

In terms of the guts of the logic, Daniel uses his knowledge of indicators - and how they can be successfully applied to produce a winning trading system - to come up with a broad logic framework out of which will fall a successful system for a given pair and history. Some of the indicators used in Coatl - and baked into its core logic – RSI, Stochastics and MACD and others.

One of the indicators used is Daniel's own proprietary indicator which is at the heart of one of his best performing systems. If you are an Asirikuy member, you probably know what i'm talking about. If you're not, you'll have to join Asirikuy to find out what is is!

The inputs used for each indicator (period length for RSI for example) are pre-set in the system and not optimized - which was quite a surprise. As an example, the system uses a 20-period and a 40-period RSI, and there are 2 RSI related rules in the system. The rules are considered separately – the system can enter on a RSI upside crossover but exit using a completely different indicator such as MACD or Stochastics. The beauty of this approach is that you don’t have to bring any traditional interpretations of indicators to the table. Daniel provides the interpretations, throws them into the mix, and out pops a (hopefully) profitable trading system.

There’s a lot more to it and here’s where it starts to get interesting. For each possible rule being considered, there is a “Normal” and “Reverse” interpretation of the indicator. So the system can go long on an upside RSI crossover, or go short on an upside RSI crossover. For each system, there are a total of 18 possible indicator configurations which can be applied.

Taking a cue from Atinalla FE, Coatl allows for 3 embedded system per pair each of which has separate entry and exit rules. Also, the system uses a built-in exit as a percentage of ATR to exit positions independent of exit signals. That’s enough information to get a broad sense of the decision space:

- For a given pair choose 3 different systems.

- For each system, choose 1 of 18 different indicator configurations for position entry.

- For each system, choose 1 of 18 different indicator configurations for position exit

- The indicator configurations consist of regular (buy on upside crossover) and reverse (sell on upside crossover) conditions

- For some indicators such as RSI and Stochastics, the indicators are used twice, once with a short-term period (20 bars) and again with a long-term period (40 bars).

- ATR-based exits cover position stop-losses, but regular position exits are covered as per the system sell logic.

So what have is a alphabet soup of trading methods that all get thrown into a huge pot and shaken throughly until the winners pop out. To use a different metaphor, you start with an entire herd of potential trading systems. Gradually the herd is culled and pared down until the best 3 systems emerge from the mass of animal flesh. When you think about it, it sure beats the traditional method of developing systems which involves examining charts and indicators and then coming up with a winner after weeks or even months of trial and error.

How does all this result in a profitable trading system? Check back for part 2 to find out

Have a great week.