In the previous articles, I have showed you the results of the automated strategy that has been running on real account for 21 months and we went through its historical backtests as well. I have mentioned at the end one important thing – the backtest itself is not enough. Why? What does it mean? And what do you need to do when a backtest is not enough?
Reasons why a backtest is not enough:
Backtest is important from several reasons, especially:
- We see how the strategy behaved in the past
- It gives us confidence in strategy itself
- It tells us if the strategy would have been working over a longer historical period and in different market cycles
Although the backtest is not insignificant, we still need something more. Finding a strategy that has good historical results is not that difficult.
It is not so important if the strategy would have been profitable in the past (this is just a must), the key is to make profit in the future. And that is something backtest cannot show us.
But how to do it? How to find out if the strategy will be profitable in the future?
We can never be sure for 100%, but we can use some pointers to help us ensure that the probability is as high as possible. We call this process robustness testing, or quality of the strategy.
How do we test the quality, or robustness?
Historical results have their specifics, we know the history, it is stable and it can not change, it cannot surprise us. While the future results are something we have no idea about, whether the market will behave in the same way, whether it will change rapidly, etc. In the future, we can expect many changes.
This is logically related to what we do with our robustness tests – we test sensitivity to change.
We test the sensitivity to change from many different angles – both sensitivity to market changes, to the broker, and strategy itself.
In other articles, I will introduce them one by one and will tell you why we are actually using them.
What we specifically do as the robustness tests?
Portfolio Test – We monitor how the strategy behaves on other markets
Timeframe test – We monitor how the strategy works on other timeframes
Slippage – how the strategy reacts to slippage
Shuffling the trades – We monitor how the change in the order of the trades impacts on the strategy
Resampling – An interesting and unique test, I will describe it in other articles
Trade skipping – how the strategy behaves when we skip for example 15% of randomly selected trades
Strategy Parameters – We monitor how strategy is sensitive to changing parameters
Volatility – We change market volatility and monitor the behavior of the strategy
3D parameter test
Strategy logic (test edge)
You can see that there are quite many tests, but do not worry, we do not have to do everything manually. StrategyQuant will help us. Many of tests I use are unique and are basically only on our platform. Most traders work with one or two tests, but this is not enough and therefore they do not have the desired resuls. Building strategies is, in short, relatively complex and I look forward to showing you how I do it in other articles.
I wish a beautiful day and look forward to another article! So far, you can read my ebook, which you find under the article.