[R-SIG-Finance] Quantstrat optimal portfolio & dynamic core-satellite strategy
Brian G. Peterson
brian at braverock.com
Wed Nov 28 01:20:43 CET 2012
On 11/27/2012 02:20 PM, Cristian Popescu wrote:
> Dear all,
> I am trying to create a dynamic core-satellite strategy and backtest it in R. So far, I think that the best solution to be able to "migrate" capital/equity from the core to a satellite and back is to create 1 account but have 2 different portfolios for the account, each with its specific symbols allocated to it:
> core.symbols <- c('SPY', 'BWX')
> satellite.symbols <-c('JNK', 'GLD')
> Thus using one of the 2 above for its designated portfolio, I think would be easier to assign symbols to the specific part of the strategy.
> My questions are:
> 1) how can i assign capital to one of the portfolios and be able to analyze and possibly change it after each day/week/month or whatever the time step will be?
> 2) how can i optimize my portfolio on a rolling time frame (tried to use fPortfolio but I need to change to classes back and forth from xts & zoo to timeSeries and back). Shoult i recreate an optimization function?
> Looking forward to your responses, and if you have any comments regarding the possible ways to do this, do not hesitate to do so.
> Thank you in advance.
I'll try to make a longer answer sometime tomorrow, but I wanted to give
you the short answer now.
The short answer is 'rebalance' rules. These currently only work on one
portfolio, but I don't think that will be a problem, as you can simply
have different rules for 'core' and 'satellite' in the single portfolio.
See the 'faber' demo for one without rebalance rules, and 'faber_rebal'
for the same strategy with a quarterly rebalance rule. As with
everything in quantstrat, it is modular, and you can write a custom
rebalance rule that will be aware of the current state of the portfolio
at the time the rebalance rule(s) is evaluated.
Brian G. Peterson
More information about the R-SIG-Finance