The impressive results have since been debunked as there were some issues with the shifts and time periods that needed adjusting. However, the results are still much better with HMM, then without. See follow up video.

The Gaussian Mixture Model (GMM) appears to carry significant weight when it comes to identifying market states or regimes that we might be in. When trying just a simple moving average crossover strategy, the GMM appears to carry a significant benefit and render not using this or something similar like an HMM as risky.

Now that Pandoras box has been opened, it is difficult not to pay attention to this and wanted to flag it up for our community asap. Just wish I had paid more attention to this sooner.

GMM's work by essentially modelling distributions and probabilities and from there, assessing what sort of hidden state the market might be in.

This is very similar to how HMM's might be used which will cover in the next video.

website: https://cryptowizards.net

Code: https://colab.research.google.com/dri...