Timing Luck

#*****************************************************************
# Load historical data
#*****************************************************************
library(SIT)
load.packages('quantmod')
tickers = spl('DBC,EEM,EWJ,GLD,ICF,IEF,IEV,RWX,TLT,VTI')

data <- new.env()
getSymbols(tickers, src = 'yahoo', from = '1970-01-01', env = data, auto.assign = T)
  extend.data.proxy(data)
  for(i in ls(data)) data[[i]] = adjustOHLC(data[[i]], use.Adjusted=T)
bt.prep(data, align='remove.na')

test.strategy <- function(data, period.ends,
	top.n = 3,      # number of momentum positions
	lookback = 250,   # length of momentum look back
	lag = 1
){ 
	prices = coredata(data$prices)  
	momentum = (mlag(prices, lag) / mlag(prices, (lookback + lag)) - 1)

	data$weight[] = NA
		data$weight[period.ends,] = ntop(momentum[period.ends,],top.n)
	data$weight
}

#*****************************************************************
# Code Strategies
#*****************************************************************
prices = data$prices

models = list()

period.ends.high = endpoints(prices, 'months')
	n.period.ends = len(period.ends.high)

n.high.in.low = 12	# i.e. there are 12 months in 1 year
period.ends.low = endpoints(prices, 'years')

period.ends.low = endpoints(prices, 'quarters')
n.high.in.low = 3	# i.e. there are 3 months in 1 quater

pe = list()
pe[[1]] = which(!is.na( match(period.ends.high, period.ends.low) ))
for (i in 2:n.high.in.low) {
	pe[[i]] = pe[[(i-1)]] + 1
	pe[[i]] = iif(pe[[i]] > n.period.ends, n.period.ends, pe[[i]])
}

#*****************************************************************
# Code Strategies
#******************************************************************    
lookbacks = c(30,120,250)
for (lookback in lookbacks) {
	weights = ifna(0 * data$weight, 0)
 for (i in 1:n.high.in.low) {
 	period.offset = i - 1  
		weight = test.strategy(data, period.ends = period.ends.high[ pe[[i]] ], lookback = lookback)

		data$weight[] = weight
		models[[paste0('Lookback_', lookback, '_Offset_', period.offset)]] = bt.run.share(data, clean.signal=F, silent=T)

			weights = weights + bt.apply.matrix(weight, ifna.prev)/n.high.in.low
		}
	data$weight[] = weights
	models[[paste0('Lookback_',lookback,'_AVG')]] = bt.run.share(data, clean.signal=T, silent=T)
}

#*****************************************************************
# Report
#******************************************************************
for (lookback in lookbacks) {
	print('Lookback', lookback)
	models1 = models[grep(paste0('_', lookback, '_'), names(models))]

	plotbt(models1, plotX = T, log = 'y', LeftMargin = 3, main = NULL)
		mtext('Cumulative Performance', side = 2, line = 1)

	print(plotbt.strategy.sidebyside(models1, make.plot=F, return.table=T))
}