How can you make data talk? Mistakes can be made

Hat tip to Jure Sah from twitter.

There are a lot of ways to make data talk through curve-fitting. yes, there is a science with the math behind these techniques for fitting, yet the choice of which technique is more art. There is no simple way to compare all the techniques, so you must make some guess on which approach is better. 

For example, with trend-following, you can use different moving averages, or you can use some form of times series forecasting. Which is better? The proof is in the returns, but those returns can be generated from the choice of the portfolio, or the risk management employed. 

At the least any curve-fitting should be compared against a few techniques to explore differences. The extra work will give anyone more confidence in the end results.