roll_backtest: Perform a backtest simulation of a trading strategy (model)...

In algoquant/HighFreq: High Frequency Time Series Management

Perform a backtest simulation of a trading strategy (model) over a vector of end points along a time series of prices.

Description

Perform a backtest simulation of a trading strategy (model) over a vector of end points along a time series of prices.

Usage

roll_backtest(
  xtsv,
  train_func,
  trade_func,
  look_back = look_forward,
  look_forward,
  endpoints = rutils::calc_endpoints(xtsv, look_forward),
  ...
)

Arguments

...	additional parameters to the functions train_func() and trade_func().
xtsv	A time series of prices, asset returns, trading volumes, and other data, in xts format.
train_func	The name of the function for training (calibrating) a forecasting model, to be applied over a rolling look-back interval.
trade_func	The name of the trading model function, to be applied over a rolling look-forward interval.
look_back	The size of the look-back interval, equal to the number of rows of data used for training the forecasting model.
look_forward	The size of the look-forward interval, equal to the number of rows of data used for trading the strategy.
endpoints	A vector of end points along the rows of the xtsv time series, given as either integers or dates.

Details

The function roll_backtest() performs a rolling backtest simulation of a trading strategy over a vector of end points. At each end point, it trains (calibrates) a forecasting model using past data taken from the xtsv time series over the look-back interval, and applies the forecasts to the trade_func() trading model, using out-of-sample future data from the look-forward interval.

The function trade_func() should simulate the trading model, and it should return a named list with at least two elements: a named vector of performance statistics, and an xts time series of out-of-sample returns. The list returned by trade_func() can also have additional elements, like the in-sample calibrated model statistics, etc.

The function roll_backtest() returns a named list containing the listv returned by function trade_func(). The list names are equal to the endpoints dates. The number of list elements is equal to the number of endpoints minus two (because the first and last end points can't be included in the backtest).

Value

An xts time series with the number of rows equal to the number of end points minus two.

Examples

## Not run: 
# Combine two time series of prices
prices <- cbind(rutils::etfenv$XLU, rutils::etfenv$XLP)
look_back <- 252
look_forward <- 22
# Define end points
endpoints <- rutils::calc_endpoints(prices, look_forward)
# Perform back-test
back_test <- roll_backtest(endpoints=endpoints,
    look_forward=look_forward,
    look_back=look_back,
    train_func = train_model,
    trade_func = trade_model,
    model_params = model_params,
    trading_params = trading_params,
    xtsv=prices)

## End(Not run)

algoquant/HighFreq documentation built on Feb. 9, 2024, 8:15 p.m.