In this video, I’ll show you one way to improve your stock market backtesting results.

This method is the one used by my research partner Roger Crandell and me when refining The 9% Signal.

One common problem with backtesting is curve fitting, which is finding a system in retrospect that would have maximized returns in the past market. For example, getting the most out of this two-year TQQQ chart:

[Chart shown in the video, at 1:09.]

The problem is that how the market moved in any period of the past is not guaranteed to repeat. What worked then, probably won’t work now.

We can extract elements from past price behavior, however, to construct a market characteristic.

For instance, this a seven-year scatter plot of TQQQ’s daily percentage price changes:

[Scatter plot shown in the video, at 3:20.]

We can defining this price behavior profile.

Then put it into a restricted randomizer that generates daily changes within this profile to simulate many markets.

This is key: Why is the real market of the past any more valid than a simulated one of the same profile?

We believe it’s not, and put the 9Sig plan through a rigorous test bed of not just a lengthy real-life backtest of 30 years, but 100 simulated 30-year markets.

Listen to this pull quote from this year’s Note 1 from The Kelly Letter:

[Quote shown in the video, at 5:42.]

From this, we zeroed in on the parameters that delivered the best results most of the time.

Could still be wrong, but offers better odds than just the one test bed of actual historical market data.