We have thoughts on this but not absolute answers. That question is part of the art of building models, and one that even we struggle with. There is no simple way to decide that we know of. We have very successful customers who regularly retrain and/or re-optimize every night whether they need to or not. Some do it every week. (Note that retraining and reoptimization are two different things of course). Our philosophy for our own trading models is to retrain/optimize only when needed, but that doesn’t mean we are right and our customers are wrong. For us “when needed” means when the model starts not working so well, which usually works out to monthly on intraday charts, or quarterly on daily charts. However, “not working so well” is a subjective concept. We wish we could give a cookbook answer but we cannot. The proof is in the pudding, so try it a couple of ways and see what works best for your objectives.
Related Articles
-
May 2021 Newsletter- Let ChaosHunter Find Inputs for Forecast Models
- 0
- 1518
-
April 2021 Newsletter – Choosing Effective Inputs for Forecast Models
- 6
- 1813
-
January 2021 Newsletter – Learn How to Build Complex Indicators – Without Programming
- 2
- 1150
-
December 2020 Newsletter – 7 Factors to Consider Before Building Your Prediction
- 4
- 1296
-
November 2020 Newsletter – Bowfort Adaptive Indicators Adjust to Market Trends
- 0
- 1284
-
September 2020 Newsletter – Use Neural Nets to Create a Co-integrated Pair Trade
- 1
- 1427