Wednesday, June 20, 2007

Currency Trading Systems

f you buy a currency trading system from a vendor, chances are it won’t make money in line with its track record. Furthermore, if you test your own in most cases it won’t produce the same in real time trading.
Why? The answer lies in curve fitting - if you don’t know what curve fitting is, read this article and it could save you a lot of money.

About curve fitting:

Curve fitting is when the system rules are bent (curve fitted) to the data, to make it produce a profit. This is very similar to shooting blindly at a barn door with a shotgun and then drawing a bulls-eye around everyone afterwards!
Curve fitting and buying a system
Most of the currency trading systems sold by vendors have great track records in back testing and in most instances never produce the gains in real time.
A hypothetical track record is exactly that:
It’s designed knowing the closing forex prices and of course it’s easy to make a profit when you know the closing price in advance.
Many vendors simply make sure the trading system makes money so they can sell it.
They know it wont work in real time, but that doesn’t matter, their after system sales.
A clue to curve fitted system is:
1. Black box – Where the logic and rules are not revealed.
2. Optimization – This is where the data has to be bent. Clues to an optimized curve fitted system are - unique rules or parameters, for different market conditions or currencies.
The Gamblers Illusion
Of course, on a few years currency trading history there is an apparent order, but just like the roulette gambler who sees number sequences repeating themselves, they will never exactly repeat in the same order again.
Bending the rules and curve fitting to make a currency trading system more profitable is a futile exercise.
If a system is robust and based upon sound logic, it should contain only a few rules and parameters and they should be applied to all currencies and all market conditions.
Of course, not all vendors selling currency trading systems deliberately curve fit, like individual traders back testing their system they do it without understanding exactly what they are doing.
Many individual traders I have seen, back test their systems and make money, but they want to improve profitability, so they simply bend the system to fit the data with lots of parameters and indicators.
When they trade it real time they lose.
They would have been better off with their simpler non curve fitted system!
No snap shot of trading history will produce the same patterns again – it’s an illusion as we have just seen.
When back testing or looking at a system just keep the following points in mind:
1. It should be based around trading the odds.
2. It should be simple with only a few rules or parameters.
3. It should trade all currencies the same way.
4. It should trade all market conditions the same way.
5. If a track record has an absence of drawdown chances are its curve fitted, so if a hypothetical track record looks to good it probably is.
Don’t fall into the trap of curve fitting!
Curve fitting is done by the majority of trading systems sold and most traders when constructing their own fall victim to it.
Curve fitting deliberate or not, is the major reason hypothetical track records that appear to give fantastic growth rates in hindsight fail miserably in real time trading.

0 Comments:

Post a Comment

Subscribe to Post Comments [Atom]

<< Home