Assume we construct a project that ensembles several rather uncorelated strategies. Each sub-strategy generates signals for different symbols and executes trades. We assign weights for each sub strategy. 

Was wondering how one would smartly assign the weights for each sub-strategy dynamically:

  • just equal weight…we dont know how each sub strategy performs and building models to predict it is not worth the effort
  • dynamically track the total return performance distribution per trailing window and assign weights accordingly. 
    • sub-strategy1: last rolling year performance contribution 10% → weight=0.1
    • sub-strategy2: last rolling year performance contribution 40% → weight=0.4
    • sub-strategy3: last rolling year performance contribution 50% → weight=0.5
  • run a portfolio optimization on trailing window, for example HPA with risk measure Sortino. We use the matrix of the the past returns of each sub-strategy
  • run an ML model to dynamically predict the ideal future weights of each sub-strategy
  • other methods?

Author