Is There Such a Thing as a CONSISTENTLY Profitable Trading System?

[lindq]

I think that much depends on how you define a system. I feel that I am as much a part of my system - if not more so - than my software and the markets I trade. It's for this reason that so many people fail with a system that another has created and profited from. How can you separate the operator from the operation? I don't think you can. At least not for any length of time.

Not to be too esoteric here, but isn't your computer and your internet connection really an extension of your thinking process into the marketplace? Aren't you, in fact, the system? Where does one end and the other begin? If you elect not to trade, or to trade for a day with different parameters, are you out of the system? Seems to me the best systems are those set basic parameters, but that accept and encourage intuition and personal involvement as conditions shift during the day, the week, or even the hour.

I'll bet for many of us who are fortunate enough to be making money at trading, the shift from losing to winning came with a shift from purely mechanical trading based on what we had learned to that point, to a deeper understanding of what the hell this is all about. We began to see the forest instead of just the trees. That change in perception doesn't happen overnight. It usually develops after a lot of pain, and many losing trades. But when it happens, I think that we become very much a part of the entire system, as well we should be.

 

[prox]

Quote from rolegario:

In my quest to define a winning trading strategy, I'm obviously worried that all my work will be wasted, and that trading is more of an art form than anything else. So, you can backtest as much as you want, but in the end there is zero bearing on your potential future profits.

Or is there? Your thoughts would be appreciated.

See if you agree with this statement:

If it makes money in a backtest, then it may or may not make money in the future.

If it isn't profitable in a backtest, then there's no way in hell it'll make profitable in the future.

That should be reason enough to give some validity in testing.

You should be testing ideas that have some inherent market logic, rather than changing around your indicator settings, moving average periods or some arbitrary stop or profit target numbers to see which ones works better.

 

[wally_]

Quote from prox:



If it isn't profitable in a backtest, then there's no way in hell it'll make profitable in the future.

No, that does not have to be true at all, it may or may not be profitable. The lack of profitability in the past does not imply the lack of it in the future pretty much as the profitability in the past does not imply the profitability in the future. You will obviously choose a system that made money in the past over another system that lost money in the past, but the reasons are purely psychological. No backtesting will ever guarantee any profitability in the future. In fact, I believe that too much time is spent on backtesting, more time should be spent on understanding why the system can work. I would not trust any system that I do not understand no matter how well it performed in the past. Of course, I would also avoid systems that lost money in the past, because I do feel more comfortable with those that were profitable in the past.

 

[prox]

Quote from wally_:



No, that does not have to be true at all, it may or may not be profitable. The lack of profitability in the past does not imply the lack of it in the future pretty much as the profitability in the past does not imply the profitability in the future. You will obviously choose a system that made money in the past over another system that lost money in the past, but the reasons are purely psychological. No backtesting will ever guarantee any profitability in the future. In fact, I believe that too much time is spent on backtesting, more time should be spent on understanding why the system can work. I would not trust any system that I do not understand no matter how well it performed in the past. Of course, I would also avoid systems that lost money in the past, because I do feel more comfortable with those that were profitable in the past.

Sure, but if you think about it.. if you were unable to produce a winning system with the benefit of hindsight, phantom fills and even logical curve fitting, then the chances of it having a chance to work in real time is abysmal.

And of course, no one would risk their money on something that produces loser after loser when they eyeball over old charts.
But you're right, past results do not imply the same results in the future regardless of profitability.

 

[TGregg]

Quote from prox:



Sure, but if you think about it.. if you were unable to produce a winning system with the benefit of hindsight and even logical curve fitting, then the chances of it having a chance to work in real time is abysmal.

And of course, no one would risk their money on something that produces loser after loser when they eyeball over old charts.
But you're right, past results do not imply the same results in the future regardless of profitability.

Hmmm, perhaps one should curvefit to maximum loss on backtesting. . .

 

[Don Bright]

To answer the question honestly: NO....there are many ways to trade for a living, but there is not a consistent winner available now or ever. The market is always adapting, and the strategies have to change with it.

Sorry....

Don

 

[PoundTheRock]

It can't be based on a single concept, i.e., it must be a framework-based system. For example, suppose I want to develop a universal opening range system. The first decision I have to make is when to identify "breakout" mode and when to identify "fade" mode. So, let's start with a couple of Crabelian concepts: narrow range days and wide range days. On any series of days, I average the ranges over the past few days and divide this average range by the average true range to obtain a ratio. If this ratio is less than 1, then we go into breakout mode (stop on either side of the opening range of say the first 30 minutes). If this ratio is greater than 1, then we fade either side of the range. The criteria do not have to be this simple; maybe we only go into breakout mode if the ratio is less than 0.8 and fade if the ratio is greater than 1.2, perhaps even dynamically adjusting the position size based on this ratio, i.e., the lower the ratio, the higher the breakout size. The key is that the system must be adaptive; I believe that Chande has a system in one of his books where he tries to distinguish between trending and non-trending.

 

[PoundTheRock]

Depends on how often you're measuring consistency, too. Show me a system that can make money every month and has plenty of losing days - great (and yes there are such systems)! Show me a new system that makes money every day for a few weeks, and I'll show you a system that's bound to have a big losing streak. I'll take the first system over the second any day.

 

[wally_]

Quote from prox:



Sure, but if you think about it.. if you were unable to produce a winning system with the benefit of hindsight, phantom fills and even logical curve fitting, then the chances of it having a chance to work in real time is abysmal.

Yeah, but curve fitting is never a good idea. I don't do it. A system should work with a broad array of values of its parameters and the fewer it has these parameters the better.

A curve-fitted system that does not make money without curve fitting is a loser to me.

 

[PoundTheRock]

Quote from wally_:



Yeah, but curve fitting is never a good idea. I don't do it. A system should work with a broad array of values of its parameters and the fewer it has these parameters the better.

A curve-fitted system that does not make money without curve fitting is a loser to me.

Yup, I prefer to see better results in the future rather than what it could have done in the past.