I have the following simple trading strategy:
Entry Signal: when the Price of IBM is above the upper Bollinger band.
Close Signal: when the Price of IBM is below the lower Bollinger band.
Here are the Bollinger Bands:
require(quantmod)
# load IBM data
tickers = c("IBM")
myEnv = new.env()
getSymbols(tickers, from="2012-01-03", to="2014-12-01", env=myEnv)
close.prices = do.call(merge, eapply(myEnv, Cl))
close.prices = close.prices[,pmatch(tickers,colnames(close.prices))]
colnames(close.prices) = c("IBM")
# extract the upper and lower bollinger band with TTR's BBands function
bb.up = BBands(close.prices, n=20, maType = SMA)[,3]
bb.dn = BBands(close.prices, n=20, maType = SMA)[,1]
The tricky part now is to close the position only if the Price of IBM is below the lower Bollinger band at the re-allocation date. Otherwise we roll the Signal of the last period to the next period. To accomplish weekly re-allocation:
# apply the startpoints function to pick the week's first trading day for
# re-allocating the portfolio
startpoints = function (x, on = "weeks", k = 1) {
head(endpoints(x, on, k) + 1, -1)
}
sig.bb.up = ifelse(close.prices > bb.up, 1, 0)
sig.bb.up = sig.bb.up[startpoints(bb.up),]
sig.bb.dn = ifelse(close.prices < bb.dn, 1, 0)
sig.bb.dn = sig.bb.dn[startpoints(bb.dn),]
The whole question now is how to define a properly coded signal function sig.bb that contains a "1" as soon as the Price is above its upper bollinger band at a re-allocation date, then holds the stock until the price is below its lower bollinger band at any following re-allocation date.
What I have tried is to catch the frist observation and then to roll all the following observations based on the first entry of the sig.bb vector
sig.bb = ifelse(index(close.prices[1,1]) == "2012-01-03", sig.bb.up,
ifelse(close.prices > bb.up, 1,
ifelse(close.prices < bb.dn, 0, lag(sig.bb))))
which returns "NA"...
How to proceed (for those interested in this topic) after sig.bb is obtained can be found here: Equally Weighted Reallocation of Stock Portfolio at Specific Dates According to a Signal
I'm not exactly sure what you want the output to be, but see if this is close
m <- merge(close.prices, BBands(close.prices, n=20, maType="SMA"))
m$sig[with(m, IBM > up) & index(m) %in% index(m)[startpoints(m, on="weeks")]] <- 1
m$sig[with(m, IBM < dn) & index(m) %in% index(m)[startpoints(m, on="weeks")]] <- 0
m$sig[1] <- 0
na.locf(m)