Change weights based on precomputed array

[al33m501]

Hi! Firstly, thank you for awesome backtesting solution.

I have precomputed weights for my portfolio, and trying to change weights every 30 days using vbt.Portfolio.from_order_func based on my precomputed array. I studied in details PortfolioOptimization.ipynb example (especially "Search and rebalance every 30 days" part), but still do not understand how to just put my weights for exact period of time in find_weights_nb. Maybe there's some info about period of time in c?(I didn't find that).
Can you give me some tip?

[polakowo]

Hi @al33m501,

specifically for the function pre_segment_func_nb, you can read about various c fields here. The current index (= row) can be accessed using c.i, while the current group (= range of columns that share the same capital) ranges between c.from_col and c.to_col. You can then do my_array[c.i, c.from_col:c.to_col] to access elements that belong to this segment, given that you built your array properly - it must be of the same shape as your price. Since in pre_sim_func_nb we defined that each 30th segment is active while all others are inactive, your c.i will be 30, 60, 90, etc.

If you want to rebalance on custom periods (for example, each 1st of the month), instead of changing segment_mask in pre_sim_func_nb, you can remove pre_sim_func_nb entirely and create your own mask using pandas and pass it as segment_mask to Portfolio.from_order_func.

The find_weights_nb function should then return the best shape ratio and an array of weights of shape c.group_len, that is, one element per column in the current group.

If you think you can build your weights array prior to the simulation, there is a simpler approach: https://vectorbt.dev/docs/portfolio/base.html#vectorbt.portfolio.base.Portfolio.from_orders

[al33m501]

Thank you! This is exactly what I was asking about.