Suggestion on DiscreteAllocation class for regular rebalancing

[HYANGKI-LEE]

Hello, Robert

I found your DiscreteAllocation.greedy_portfolio() in pypfopt while implementing a regular rebalancing task. However, my challenge was that my situation was not SIMPLY convert continuous weights to discrete allocation.

A new situation I had to consider was:

• I already have some stocks.
  • In other words, we had to consider the situation of not only buying the stock, but also selling it.
• ‘fee’ (a kind of contraction cost) must be considered (In addition, using ‘total_portfolio_value’, which is dependent on ‘latest_prices’, felt a little less intuitive to me.)

Thankfully, thanks to your project, I was able to implement the task without difficulty. I really appreciate this part.

I am using your code by modifying the structure as below.

[Inputs & Outputs in DiscreteAllocation class] “””

• Inputs:

  • weights - dict
  • latest_prices - pd.Series or dict
  • total_portfolio_value - int/float
  • short_ratio- float

• Output:

  • allocation - dict “””

[After modifying] “””

• Inputs:

  • weights - dict
  • init_state - dict # number of existing shares
  • latest_prices - pd.Series or dict
  • cash - int/float # amount of cash currently has
  • fee_rate - float # fee rate as contraction cost
  • short_ratio- float
  • verbose- bool

• Output:

  • allocation - dict
  • action - dict # action for desired weight (+ indicates buying, - indicates selling)
  • cash_left - dict “””

If this fixing fits the direction of your project PyPortfolioOpt, I'd be happy to share with you.

Please read the above and reply. Thank you for your time.

[robertmartin8]

If this can be implemented in a self-contained way with a couple of unit tests, I'm happy to accept a PR!

[AnkitAggarwalAlphagrep]

@HYANGKI-LEE Can you please create a PR, I would like to use that code.

Thanks