Portfolio

PortfolioAtDate

A portfolio at one single instance in time is represented by a PortfolioAtDate object. It contains the information about the quantity of all holdings and cash at that date.

Structure

struct PortfolioAtDate
    holdings::Vector{HoldingAtDate}
    cash::Float64
    short_sell_cash::Float64
    margin_cash::Float64
    date::DateTime
end

• holdings: Vector of individual asset holdings. See HoldingAtDate for details.
• Cash positions:
  • cash: Regular cash balance available for trading
  • short_sell_cash: Cash generated from short-selling activities (always equal to the sum of all negative holdings, i.e. all short positions)
  • margin_cash: Cash frozen as a margin in case of margin requirements for short selling.
• date: Date and time of the portfolio snapshot

If margin requirements are 150% and holdings for a total of $100 are shorted then short_sell_cash will contain $100 and margin_cash will contain $50.

Key Functions

• get_tickers(pf::PortfolioAtDate): Vector of tickers for all holdings
• get_quantities(pf::PortfolioAtDate): Vector of quantities (holdings + 3 cash positions)
• get_prices(pf::PortfolioAtDate): Vector of prices for holdings + 3 cash positions as 1.0.
• get_values(pf::PortfolioAtDate): Vector of values ($\text{price} * \text{quantity}$) for holdings + 3 cash positions.
• get_total_value(pf::PortfolioAtDate): Total portfolio value

HoldingAtDate

Represents a single asset held in a portfolio at a specific point in time.

Key Functions

• get_ticker(holding::HoldingAtDate): Returns the ticker of the given holding at the given date.
• get_price(holding::HoldingAtDate): Returns the price of the given holding at the given date.
• get_quantity(holding::HoldingAtDate): Returns the quantity of the given holding at the given date.
• get_value(holding::HoldingAtDate): Returns the value (quantity * price) of the given holding at the given date.

Setup of PortfolioAtDate

To setup an initial PortfolioAtDate from a Collection we use the initialize_portfolio() function. The function takes a Collection and a datetime and creates a PortfolioAtDate at that datetime with all the Assets in the Collection available as holdings. It also takes a cash argument which will set the amount of cash available. All holding qunatities are set to 0.

PortfolioHistory

A PortfolioHistory represents the evolution of a portfolio over time, containing snapshots of the portfolio before and after reweighting at each trading date.

Structure

struct PortfolioHistory
    name::String
    pf_before_rw::Vector{PortfolioAtDate}
    pf_after_rw::Vector{PortfolioAtDate}
end

• name: A string identifier for the portfolio history.
• pf_before_rw: Vector of PortfolioAtDate objects representing the portfolio before reweighting at each trading date.
• pf_after_rw: Vector of PortfolioAtDate objects representing the portfolio after reweighting at each trading date.

Key Functions

• get_total_values(pfh): Vector of total portfolio values before reweighting
• get_dates(pfh): Vector of dates corresponding to each portfolio snapshot
• get_number_of_holdings(pfh, ind; cutoff=0.0000001): Number of holdings at a specific index with quantity above the cutoff
• compound_returns(pfh): Vector of cumulative returns relative to the initial portfolio value
• returns(pfh): Vector of period-to-period returns
• returns_log(pfh): Vector of log returns

Portfolio Reweighting

At every timestep new portfolio weightings are calculated from PositionScores.

The final portfolio weights for each asset $i$ at timestep $k$ are calculated as:

\[W_{i, k} = \frac{S_{i, k}}{\sum_{i} S_{i, k} + S_{c,k}}\]

where: $S_{i, k}$ is the score of holding $i$ at timestep $k$ and $S_{c,k}$ is the cash score at timestep $k$

The cash weight is calculated similarly:

\[W_{c, k} = \frac{S_{c, k}}{\sum_{i} S_{i, k} + S_{c,k}}\]

The sum of all weights (including cash) always equals 1:

\[\sum_{i} W_{i, k} + W_{c, k} = 1\]

Setup And Propagation of PortfolioHistory

A PortfolioHistory can be created in two main ways:

1. From a PortfolioAtDate: Use the propagate_portfolio() function, which takes an initial PortfolioAtDate, a target date, and a PositionScores object.

2. Directly from PositionScores: Use the PortfolioHistory() constructor, which creates a PortfolioHistory with all holdings from the PositionScores available.

can also be further propagated... write more

PortfolioHistoryDaily

A PortfolioHistoryDaily represents a collection of daily PortfolioHistory objects, allowing for a more granular view of portfolio changes within each day.

Usually used in conjecture with Discrete Collection

Structure

struct PortfolioHistoryDaily
    name::String
    pfh_days::Vector{PortfolioHistory}
end

• name: A string identifier for the daily portfolio history
• pfh_days: Vector of PortfolioHistory objects, each representing a single day's portfolio changes

Key Functions

• get_total_values(phd): Vector of total portfolio values at the end of each day
• get_dates(phd): Vector of dates corresponding to each day's PortfolioHistory
• returns(phd): Vector of daily returns
• returns_log(phd): Vector of daily log returns

PortfolioHistory vs PortfolioHistoryDaily

1. Use Cases:

   • PortfolioHistory: Suitable for end-of-day rebalancing strategies.
   • PortfolioHistoryDaily: Ideal for intraday trading or when tracking intraday events is neede.

2. Data Access:

   • PortfolioHistory: Access via pf_before_rw[i] or pf_after_rw[i] for i-th trading date.
   • PortfolioHistoryDaily: Access via pfh_days[i] for i-th day, then use PortfolioHistory methods.

Analysis of PortfolioHistory

The following mostly self explanatory functions are useful when evaluationg a trading strategy:

sharpe_ratio(pfh::PortfolioHistory)

maximum_drawdown(pfh::PortfolioHistory)

plot_portfolio(pfh::PortfolioHistory)

plot_portfolio(pfh::PortfolioHistory, pfh2::PortfolioHistory)

plot_portfolio(pfh_vect::Vector{PortfolioHistory})

print_portfolio_table(pfh::PortfolioHistory

Most of these functions also work on PortfolioHistoryDaily.

For all functions see Portfolio Analysis - Functions.