Trading Strategy Services
Trading Strategy Analysis services, backtesting, optimisation and historical walk
forwards testing
Backtesting
Generally speaking, when developing a trading system, only 20% of the job should
involve developing the trading system itself — developing formulas, designing indicators,
and so on. Now, 80% of the job should involve validating the trading system, proving
that the signals it provides lead to profitable trading decisions. Backtesting using
historical data and forward-testing should be part of the validation process of
any trading system. Think of it as parallel to developing laser-guided missiles
where the testing, calibration, and global positioning systems have the same importance
as the laser guidance system itself.
Why is backtesting important?
Backtesting is the first and easiest way to validate the signal algorithm. How can
anyone trust colored bars or lights or whatever indications you are using if you
have not validated it by backtesting? If you blindly follow unproven signals, you
are just shooting in the dark. It will be costly. So make sure that whatever trading
software you are using offers backtesting capabilities. You can validate the signals
and strategies before you risk any hard-earned money. Trading with proven strategies
is the best way humanly possible to trade in the market.
What is the importance of forward-testing and paper-trading?
Backtesting involves looking at data from the past and analyzing the performance
results. Forward-testing should really be called historical forward-testing. It
doesn’t mean testing from here on out. Assume we are looking at data from six years
ago. Since we are now in 2009, it would mean going back to 2003, at which time we
set the parameters of the program. So we test the system in 2004 and see how it
worked. Then we run the program again at the end of 2004, fix the parameters, and
then test it to see how it performs during 2005. So you keep forward-testing in
this way. The advantage of doing this is it gives more realistic results.
What about virtual paper-trading?
When you are paper-trading — that is, trading your system without risking any capital
— you will buy and sell based on the buy or sell signals on the chart. The program
will log in all your buys and sells. So you let the program go back, say one year,
and then test it. This is a great way of testing to see if your system gives you
the desired results.
When you test and analyze a trading system, what kind of variable do you consider?
For starters, we suggest you not change anything in your trading system when you
are trading it virtually. If you do, the entire program will change so you don’t
know if it will work. So when paper-trading, just keep your parameters fixed. Let
the program go back one year and let it run so you can see how the program works.
You may see that half your trades win and half lose. But as long as your losses
are small and your wins are big, you could end up having four to five good trades
a month. This is why trend-following works.
What are the disadvantages of curvefitting?
Curve-fitting tests how well the model is working for the past data. It has nothing
to do with the future. Since we are trading the current and future markets, in most
cases, curve-fitting parameters may not work for future markets. The more input
parameters you use, the better the fitting can be. But since you are using more
parameters, it becomes much more difficult to fit future market conditions.
Source John Wang
Technical Analysis of Stocks & Commodities