Handling variable market conditions.

 

I've been playing around and experimenting with various trading systems for some time now

I've only ever succeeded in designing a system around one of those market conditions, but never both...

1. Trend based systems (e.g Moving Average based systems) work well in trending markets, but drawdowns come thick and fast in a ranging market as you get whipsawed in and out.

2. Ranging systems (e.g the grid scalping type systems, which rely on the market to come back OR oscillator based systems) are lucrative in range bound markets, but once the market begins to trend strongly, they too result in quite severe draw downs.

...and I might as well include in here a third market type, 3. Vertical (where price sensitive news is annouced as someone suggested in another forum)

The main angst is that I've never been able to sufficiently detect in a timely matter (before profits get significantly eaten into), when market conditions for a particular system, (the system is not fit for current market conditions) change and when I should stay out

and vice versa...

I've always had trouble detecting in a timely matter (before a significant opportunity to profit is lost) when market conditions are now ripe and when I should get back in.

Does anyone include with their suite of trading systems a set of rules to determine when to either stay out -or- when to dive into your tool box and grab another system more suited for current market conditions to trade?

Or do people have up their sleaves, another method to mitigate this situation, for example, by combining multiple trading systems?

Using a suite of multiple trading methodologies together is something I'm presently investigating, but will admit that both Manual and automated Back testing results have so far been rather lack lustre in terms of the returns I could (historically) get as oposed to other Investment vehicles.

 

To no avail, the Holy grail

 

Neural Networks works as a brain evaluates circumstances. A flow chart of logic has to be created to cover all the what-if's conditions, including validating crossovers. This is where vast, if not 99% of all the EA's fail and will continue to fail. The path to the holy grail of currency investing is expanding the simple logic of the EA's into a formidable conditional logic neural network. You have realized what I have realized about these forums. I have started to rethink the 2 ea's that I am working with to incorporate more neural logic sequences to resolve most market conditions. If you know flow diagramming, I would suggest you test it in the flow diagram before making the ea. It has to cover all the conditions you have stated, and a flat market,

I lost $120.00 Monday on the flat market because I failed to install safeguards for this condition. I am making modifications to my moving average ea to minimize this in my program in the next few days. Good luck in rethinking your ea!

Dave

<<

 

I would like to also state for the record that it would help us all out if someone could finally make a backtester that works at least 98% accurate instead of this sorry tester we waste our time with, or have given up on. This would increase our efficiency of design development by about 1000%.

I am not a mq4 programmer, but it astounds me why someone has not solved this problem in 2 1/2 years, extending from mt3 to mt4 ea's. What good is a marginal backtester, especially considering that 90% quality model is still marginal quality. Would you fly a plane or rocket with only a 90% guarantee of flight success if you had your family and children on it and their future at stake?? Then why doesn't somebody work to create a highly reliable backtester or fix the existing one???? How many people on these forums must keep on asking this very question over and over? - why doesn't somebody fix the thing or create a new one?

I know there is another system out there to create your own ea and test it, but that is not the solution needed. A backtester needs to run all ea's that can run on the existing mt4 platform, but be at least 98% accurate. Is it that Meta Organization is in denial or are they too cheap or not competent enough to fix a flawed backtester?? I surely would like a straight answer for once or even better, an answer saying it is being fixed once and for all! What is the problem with solving this now??

Dave

 
iscuba11:
I would like to also state for the record that it would help us all out if someone could finally make a backtester that works at least 98% accurate instead of this sorry tester we waste our time with, or have given up on. This would increase our efficiency of design development by about 1000%.

I am not a mq4 programmer, but it astounds me why someone has not solved this problem in 2 1/2 years, extending from mt3 to mt4 ea's. What good is a marginal backtester, especially considering that 90% quality model is still marginal quality. Would you fly a plane or rocket with only a 90% guarantee of flight success if you had your family and children on it and their future at stake?? Then why doesn't somebody work to create a highly reliable backtester or fix the existing one???? How many people on these forums must keep on asking this very question over and over? - why doesn't somebody fix the thing or create a new one?

I know there is another system out there to create your own ea and test it, but that is not the solution needed. A backtester needs to run all ea's that can run on the existing mt4 platform, but be at least 98% accurate. Is it that Meta Organization is in denial or are they too cheap or not competent enough to fix a flawed backtester?? I surely would like a straight answer for once or even better, an answer saying it is being fixed once and for all! What is the problem with solving this now??

Dave

If you know exactly how your strategy/system works, you shouldn't need any backtesting, but rather perform live testing trading paper money. That has worked for me very well personally. Those who don't test their strategies in real time fail, mainly because they have no idea how a specific strategy works, or are just lazy.

 

If I had a crystal ball that could tell me all the conditions that can happen in the market, then I would not have to backtest. Unfortunately, this is rather unrealistic at best. This is why backtesting is crucial to the long term success of an EA. To put a stamp of 100% guarantee on a few weeks or even a few months of forward testing is like gambling at the casino. Not only does it waste weeks and months, but will not guarantee the performance over many different market conditions. This is why most of the ea's fail over time because the programmer did not create logic conditions to filter all variations of market conditions. What the backtester could reveal in minutes will take years to reveal in forward testing.

That is one reason computers were made in the first place. Only if forex programmers would create a sound backtester, or originators of the backtester would fix their next to worthless backtesting software to work right, once and for all, will this ever be achieved. If the backtester worked right with a 98% accuracy, I could evaluate 'every existing ea' in one day, and then see what corrections would do, and have it tested to see the overall end result in minutes, not weeks.

If a backtester works on backtesting with 98 - 100% accuracy through various market conditions, then you have a sound EA, and not before. Why does everybody seem to live in denial or refuse to try to fix the strategy tester? This totally astounds me in this day and age. Is it that hard to make a backtester that works the same way a trading platform works, with the same accuracy as the trading platform?? It seems we are still living in the stone age and not the computer age?? Has anybody ever heard of reverse engineering before? If it created results going forward, why can it not create the same results going backwards?? Or replaying the same raw data during the same timeframe using the same data input stream as the trading platform used and you should achieve 100% accurancy. Why the heck is this not being done?? Hard drive space is cheap - What is the problem? Even creating a CD or DVD packed full of raw data the same form as the trade platform receives it is no big deal - What is the problem??

 

To follow on from my original post, From much of my testing to date, it appears that purely mechanical systems when run over a long time frame never seem to get more than lack lustre gains. The problem is at the moment I've personally never been able to see how one could logically program into an E.A to detect when a market's behaviour changes, including using a magnitude of "what if" statements.

It is looking more and more apparently to me that a discretionary element has to be applied and where imperical decisions must be taken from time to time by drawing on their own experience from years and years of Market experience if one would get more than lack lustre results.