How long should you test a strategy?

Hi All,

I wanted to chime in here with my take...

First of all lets be clear on something... yes static systems will fail you eventually but that is not to say the approach will fail. What I mean but this is that if your stops and targets etc are just arbitrary figures like 10 pips all the time no matter what then yes when volaitility increases you will get whipped out only to see the position eventually go on to the target.

But if for example like me you have an approach which encompasses dynamic stops and targets using support and resustance and swing points then you should be good to go for the long run and you can test this.

Why is this? well if you are placing you stops behind a recent swing then you will automatically take the market volatility into account and if your targets are at a support or resistance zone then you will also be taking the market dynamic into account.

Past this you need to understand the market conditions... trending, consolidating or reversing. Once you have all of this you will be onto a winner.

When I first started trading and people told me that I will always have to change my system I got really worried and thought I can't keep going through this process, but then when I understood that actually I can have the same philosophy and just adjust stops and targets etc I was ok.

Then what ever you are testing make sure it is goes through all periods... consilidation, trending etc to see how it reacts to them. Note that it won't perform well in all conditions so don't expect this, you just need to come out net profitable. More importantly don't cheat yourself, take all setups even the losing streaks.

Finally when it comes to live testing keep a detailed journal and keep updated performance records so as well as your own trading record go pver every day and trade in hindsight to see how you fair against th eactual rules as we all know the traders screw themselves in the end.

Just my 2 cents.

Nexas

The question was if the market through the years had change its behavior or not
And I believe it did change especially the last years that volatility
Explode .
And I just give a example with Donchian channels which is not the only one of course even consolidation strategies Like Pivot trades, was more clear back then
Look at your charts as well how the body’s of the bars was 5-10 years ago and how is on our days

For me, it's not about "how long".

It's about "how many trades".

You should test until the results are statistically significant within the limits you've pre-defined for yourself, according to how risk-averse you are.

It's a statistical function: you need to work out the standard deviations in the results, and so on.

For example, if you're looking at a system with a 50% strike-rate of positive results, which always makes the same profit on the winning trades and/or the same (smaller!) loss on the losing trades, you won't need to analyze an enormous number of trades to achieve a high degree of statistical significance, because one or two trades "going the other way" out of 100 trades (say) isn't typically going to make an overwhelming difference.

On the other hand, if you're looking at a method which (for example) has only one winning trade in four, winning 80 pips, and the other three trades in four losing an average of 15 pips each, it certainly sounds profitable overall, but you're going to need a far bigger sample size to be sure about that, because with a successful strike-rate of only 25%, you'll need to measure a lot more to be confident of the figures (and you'll also need to allow for a "longest losing run" of at least 35 trades, which you'll encounter at some point - and a "longest losing patch" which is even longer than that).

The further away the "strike-rate" is from 50/50, the bigger the sample you'll need, for the observed results to be predictive. (But not on a linear scale).

In reality, with many systems, it's much more complicated than this allows for, because the size of the profits from profitable trades can vary so much.