← Regimes Hypothesis Implementation

Theory / Hypothesis

We suppose that coin statistical distributions (e.g. std and mean) change on average when a certain boundary is crossed.

Example: If for a period of time the coin’s std is 2-3x greater than historical mean, we are in a regime where mean and std would be different.

Hypothesis: The correlation to features and types of features change - which could be checked by silhouette score.

Process

  1. Sample 50 coins
  2. Find a discriminator for which statistical properties alter
  3. Test on other part of data to check they alter sufficiently there as well
  4. Measure if it’s a good generalization

After features are received:

  1. Use the measure to separate data into regimes on sample of coins and data
  2. Run silhouette to verify separation by the change of correlation to the features