What do you think?

Hello guys, I've elaborated some trading strategies for different stocks, and I have backtested them over the last 11 years. The results are very compelling. The problem is that when I test them over another period (1997-2007), the results get materially different and worse. What do you think about this problem ? What do you think would be the best way for proceeding ? Thank you for any advice.

 

And where did u elaborate on these strategies ? Bc I searched the forums and came up with nothing.

 

You think I m gonna pubblish a several months work ? Gotta be kidding. Just would like to know your opinion if you want to share. Very welcome from some expert trader with more than 5 years experience

 

Okay. You're new.

You posted a question about something you know so little about, you don't even know the vastly important cues/clues you left out.

But as it stands, your question is unanswerable. (Although, 99% of those reading this thread have no doubt what the answer *probably* is.)

To save a lot of bother, look at a 20 year chart. Draw a line down the middle. Look. Look again.

 

If you'd be more specific I d appreciate. Thank u

 

He was specific. Go pull up a 20 year chart of an index. Use your critical thinking skills as you look at the chart. Break it down to your time periods. Take a look what happened. Not that difficult.

 

Run a regression analysis using 5 to 6 factors you believe can predict the direction of interest rates and run that regression over interest rate data from the 1980s. Then run the same regression on data from the 1990s using the same 5 to 6 factors. Compare the regression outputs and see how inputs and correlations are not stationary.

 

knew that covariance is not stationary, but thank u. I was more interested to know what do you do in this case. How do you operate. If you want to say of course

 

I mean , the result that I got are very nice over 2007-18, but I am a bit scared by what would have happened with the same strategy over the 97-07. SO i am wondering if I should try to find something that works over the entire 20 years period.... of course, because covariance is not stationary probably considering another 20 y period would give different results......