In sample vs out of sample results

The vast majority of strategies that I generate works beautifully in the “in sample” period, but doesn’t work at all in the out of sample period (see attached example). Does this just mean that the strategy is not effective in today’s market, or am I doing something wrong in my testing?

Hi,

this is absolutely normal that 99.99% strategies fail when tested using OOS data. Only those that prove to be useful on OOS data are meant to be further investigated as potential trading strategies

How about TF ?

What is the optimal “time frame” for QSX ? Is it possible to calculate a reasonable strategy for 15M ? Or should one focus on longer TFs ?

For several months I have been trying to calculate 15M for EW30, initial selection on 4H, and then further tests on 15M. Unfortunately, the results are very poor and transferring directly to, for example, SP500 is a disaster.
So how is it with this TF ?

Regards

Kris

Hi,

I personally prefer top-down approach when it comes to selecting the right timeframe. Higher timeframes tend to look easier to search good strategies for

Higher TF you mean 1H, 4H, 1D etc.

lower: 1Min, 5Min, 15Min

Am I right?

Regards

Kris