In the current package version, one can specify every how many days the portfolio needs to be rebalanced/reoptimized (with the arguments optimize_every and rebalance_every. However, one may want to specify a different rebalancing schedule, like every Monday or every last day of the month.

This feature will be added, hopefully in a way that the interface is still clean and simple.

Include the transaction costs in the backtesting (in the current version they are ignored).

This future feature is mentioned in the vignette:
https://cran.r-project.org/web/packages/portfolioBacktest/vignettes/PortfolioBacktest.html#future-features

The wealth series calculated by portfolioBacktest() looks odd to me.

I have a simple portfolio function which always returns a weight of 1.
I call portfolioBacktest() with the following parameters:

bt <- portfolioBacktest(portfolio_funs = list("rwf-long-only" = rwf_long_only),
, dataset_list = list(data_bt) # Enclose flat dataset in list()
, price_name = "close"
, show_progress_bar = TRUE
, shortselling = F
, lookback = 13
, optimize_every = 1
, rebalance_every = 1
, execution = "same period" #c("same period", "next period")
, cost = list(buy = 0, sell = 0, short = 0, long_leverage = 0)
)

Here are the results:

t1=cbind(bt$"rwf-long-only"$data1$w_rebalanced, bt$"rwf-long-only"$data1$X_lin, bt$"rwf-long-only"$data1$w_bop, bt$"rwf-long-only"$data1$wealth)
setlastcolnames(t1, c("w_rebalanced", "X_lin", "w_bop", "wealth"))
w_rebalanced X_lin w_bop wealth
2022-02-23 NA NA NA NA
2022-02-24 NA 0.015049063 NA NA
2022-02-25 NA 0.022064003 NA NA
2022-02-28 NA -0.002558613 NA NA
2022-03-01 NA -0.015230200 NA NA
2022-03-02 NA 0.018396181 NA NA
2022-03-03 NA -0.004978448 NA NA
2022-03-04 NA -0.008124677 NA NA
2022-03-07 NA -0.029479220 NA NA
2022-03-08 NA -0.007581580 NA NA
2022-03-09 NA 0.026810758 NA NA
2022-03-10 NA -0.004515496 NA NA
2022-03-11 1 -0.012715083 NA 1.0000000
2022-03-14 1 -0.007308275 1 0.9926917
2022-03-15 1 0.021990454 1 1.0145215
2022-03-16 NA 0.022174131 1 1.0370176

I think wealth[x] should be: wealth[x-1] + (w_bop[x] * X_lin[x]).

So i would expect the wealth series to look like:
1.0000000
0.9926917 (1.0000000 + (1 * -0.007308275))
1.014682 (0.9926917 + (1 * 0.021990454))
1.036856 (1.014682 + (1 * 0.022174131))

But the last 2 values are different in the wealth series produced by portfolioBacktest()!

What am i missing here?

plotPerformanceVsParams(bt) throws the error: No portfolio found in backtest!

In the code for this function on line 7 (N_portfolios <- length(attr(bt_all_portfolios, "portfolio_index"))) it is looking for an attribute called "portfolio_index" in the backtest object, but there is none.

I am using the portfolioBacktest package which is on CRAN.

It would be a great help if it was possible to save/access state inside portfolio functions, especially for more complex portfolio functions.

Using the .GlobalEnv with assign() does not seem to work for this purpose, even when setting the paral_datasets parameter.

If the user-defined portfolio function could receive the currently held portfolio w0, then it would allow a greater flexibility. For example, one could do:

• control the turnover in the portfolio design
• depending on the turnover, decide to rebalance or not

This is mentioned in the vignette as a future feature:
https://cran.r-project.org/web/packages/portfolioBacktest/vignettes/PortfolioBacktest.html#future-features

Hi Team,

Using the function "backtestSummary", Could you advise if the annual return accounts whether daily or weekly or monthly data is used?
There is no visibility of how it is calculated and it is not mentioned in the document.

If i use the daily data points versus if i were to use weekly data points, does the annual return take into account if the frequency of the data?

Please advise

I would like to backtest simple momentum strategy. I would like to buy 10% osf stocks with highest mmentum (say half year return).

I would like to rebalance every month (buy on the begining of the month).

Here is my try:

data("dataset10")
quintile_portfolio_fun <- function(dataset, w_current) {
  X <- diff(log(dataset$adjusted))[-1]  # compute log returns
  N <- ncol(X)
  # design quintile portfolio
  ranking <- sort(colMeans(X), decreasing = TRUE, index.return = TRUE)$ix
  w <- rep(0, N)
  w[ranking[1:round(N/5)]] <- 1/round(N/5)
  return(w)
}
portfolios <- list("Quintile"  = quintile_portfolio_fun)
bt <- portfolioBacktest(portfolios, 
                        my_dataset_list, 
                        benchmark = c("1/N", "index"), 
                        lookback = 252 / 2, 
                        rebalance_every = 22)

but I get an error:

Backtesting 1 portfolios over 5 datasets (periodicity = daily data)...
Error in (function (portfolio_fun, data, price_name, shortselling, leverage,  : 
  The reoptimization period has to be a multiple of the rebalancing period.

I set rebalance_every = 22 because I wnt monthly rebalancing and I set ookback = 252 / 2 because I want to look half year momentum. Or should I set half year inside quintile_portfolio_fun function, something like:

ranking <- sort(colMeans(tail(X, 22)), decreasing = TRUE, index.return = TRUE)$ix

Finally there is some activity on backtesting in R community :) There are multiplie options in python but not so many in R.

I am aware your package is more about portfloio backtesting, but I am curious is there a simple way to backtetst strategy in which I hold / not hold one asset (0 and 1 without short selling). So I have a vector of prices on 1 minute basis and I know dates on which I hold the position. I would need the performance, expecialy relative performance (relative to banschmark).

Hi,

I have access to database returns, fundamentals and estimates. Could you show how to import data into the relevant format for the backtest in a vignette or a file?

In the current package version, the rebalancing parameters are common for all the list of portfolio functions to be backtested. However, it may be interesting to be able to specify different parameters for each strategy function. One example of use would be to try the same strategy but for different rebalancing frequencies.

This feature will be included, hopefully in a nice way that doesn't destroy the simplicity of the interface.

Allow download data from a local database. Either directly in stockDataDownload() or as another function. This will be great for speed (only download once) and also so that one can clean and curate the data once and then just load.

Provide more performance measures than the ones currently provided. For example, the Sortino ratio.

This future feature is mentioned in the vignette:
https://cran.r-project.org/web/packages/portfolioBacktest/vignettes/PortfolioBacktest.html#future-features