Backtesting - whats the point exactly?

[Yamato]

In order to know if backtesting makes sense or not, let's ask ourselves a simple question - why do I think my method will work?

The simple answer is - because I think it has worked in the past.

Backtesting tells you if it has indeed worked in the past.

Now, the problem is that most of us are so bad at programming, that even a good system, once we have coded it, turns out to be unprofitable. Yet this is not a good reason to discard backtesting.

 

[SOCRATES]

In order to know if backtesting makes sense or not, let's ask ourselves a simple question - why do I think my method will work?

The simple answer is - because I think it has worked in the past.

Backtesting tells you if it has indeed worked in the past.

Now, the problem is that most of us are so bad at programming, that even a good system, once we have coded it, turns out to be unprofitable. Yet this is not a good reason to discard backtesting.

So, from your point of view, what would be a good reason to discard backtesting then ?

 

[Yamato]

I think a good reason to set it aside temporarily, without saying "backtesting is worthless", is if you have a method that works, and don't have time to code it properly.

I am not saying that writing trading systems is the only way to become profitable, and maybe it's not even the easiest way.

But we can't say that backtesting a system is useless, just as we don't say it's not important how the market behaved in the past. That's actually the only thing we can use for our trading - we know how it behaved in the past.

What is useless is trying to code and backtest a system, if we don't have enough time and energy and knowledge to do it properly.

 

[Trader333]

I know for a fact that it is not possible to code how I trade and it never will be.


Paul

 

[StockCoach]

Backtesting is simply gathering data so that you can have something to compare your results to once you begin trading. It is a sort of guidepost that lets you know stable the ground is that you are trading on.

Of course there are a few good traders who trade without backtesting (although they do some sort of backtesting in their heads), but for the majority of traders, a systemized approach to the markets is the path to success. This is because most humans cannot control their emotions, plain and simple.

You must be able to trade in a consistant manner, if you wish to be able to control your results.

 

[zab]

Backtesting - whats the point exactly?

So that you can test all the strategies listed in this forum and see what crap they are.



Everyone backtests. It's just that some people use a computer, some don't.

Not using a computer means you can look at a wider range of strategies and be more flexible in your approach. But it is also slower and more prone to errors. You start by reading all the contradictory junk around in trading books and on bulletin boards and you have no idea what works and what doesn't. By using a computer you can quickly sort the wheat from the chaff.

The problem with using a computer to backtest is that people think all they need is to find the 'right' inputs to an MA, Stochastic and RSI and they will become rich. It's a lot harder (and also a lot simpler) than that. There are plenty of people who run mechanical systems very profitably but they do it with properly designed and tested systems. And they run with portfolios of systems and on portfolios of markets. The main advantage (IMHO) of mechanical systems is their use of money management techniques to dramatically increase returns.

The people who say they don't backtest are people who have either been around a long time (so have done a lot of forward testing), have learnt from other people (so have had others backtest their system for them) or have learnt so much from backtesting that they now know how to trade (even if they don't currently use backtesting).

Horses for courses.

 

[zentrader22]

@all,

if you trade based on mechanical trading systems, backtesting is an important step, but it's not sufficient to trade real...

My two cents:

1. i extend my backtesting of the original data with a simulated additional testing of "synthetic" data, which are generated based on some properties of the original data. If there also good results with this simulated data i'm more confident, that the system may be profitable...

2. i simulate my test results (system report) with additional monte carlo simulation methods.

For example: if i have a backtest of an EOD system which has 250 trades on 1250 bars (5 years of trading), i test this 5 year period e.g. for another 5.000 periods (> 1 mio trades) and see possible profits and drawdowns...

So I can see, if I am comfortable with the system or if I have a bad feeling...

It's also not a riskfree approach - but it give's a deeper view of the quality of the trading system!

If you are interested you can download, some freeware here:
http://www.zentrader.de/download.html

bye,
zentrader