Hi everyone, I am trying to test a strategy with random parameters that change every month as some kind of a benchmark to compare my other strategies against. This strategy is applied to the same instrument every month. I tried to make a loop with applystrategy(), where in each iteration the different random parameter values are passed, and the next month's data is used. The first iteration is executed without problems but I cannot make the next iterations work. I want to be able to use chart.Posn(), tradeStats() (in consolidated form), chart.ME(), getOrderBook(), getTxns() with the consolidated strategy backtest. How should I go about doing this? No code is necessary for an answer, but I appreciate it if there is. Just outline/plan of the code is sufficient. Thank you.
Consolidating Backtests
3 messages · Akane Fortuna, Brian G. Peterson, Ilya Kipnis
On Sun, 2015-08-16 at 20:19 +0300, Akane Fortuna wrote:
I am trying to test a strategy with random parameters that change every month as some kind of a benchmark to compare my other strategies against. This strategy is applied to the same instrument every month. I tried to make a loop with applystrategy(), where in each iteration the different random parameter values are passed, and the next month's data is used. The first iteration is executed without problems but I cannot make the next iterations work. I want to be able to use chart.Posn(), tradeStats() (in consolidated form), chart.ME(), getOrderBook(), getTxns() with the consolidated strategy backtest. How should I go about doing this? No code is necessary for an answer, but I appreciate it if there is. Just outline/plan of the code is sufficient. Thank you.
The challenge with what you're trying to do is that the indicators and signals are presumed to be vectorized functions. So they are applied to the entire dataset at once. You're talking about subsetting the data by months, and applying different parameter values. If your indicators and signals don't have any lookback period, or are constant over the entire test, then you should be able to use a rebalance rule, which will subset the data and let you change things around on your rebalance periods. Assuming that's not the case, your options are somewhat more limited. You have no problems using the same portfolio and account over progressively increasing time periods. You're basically just adding new transactions over time. The slightly challenging point will be in controlling the start and end periods for everything else. You'll need to make sure you don't have signals before the start of the subset period, or after the end of the subset period. A minimal reproducible example would help others help you. Regards, Brian
Brian G. Peterson http://braverock.com/brian/ Ph: 773-459-4973 IM: bgpbraverock
Set a seed. Loop through months, using that seed to generate the integer value of your lookback every month. Loop through all unique integer lookback values. Compute your indicator for those values. Proceed as usual. On Sun, Aug 16, 2015 at 3:02 PM, Brian G. Peterson <brian at braverock.com> wrote:
On Sun, 2015-08-16 at 20:19 +0300, Akane Fortuna wrote:
I am trying to test a strategy with random parameters that change every month as some kind of a benchmark to compare my other strategies against. This strategy is applied to the same instrument every month. I tried to make a loop with applystrategy(), where in each iteration the different random parameter values are passed, and the next month's data is used.
The
first iteration is executed without problems but I cannot make the next iterations work. I want to be able to use chart.Posn(), tradeStats() (in consolidated
form),
chart.ME(), getOrderBook(), getTxns() with the consolidated strategy backtest. How should I go about doing this? No code is necessary for an answer,
but I
appreciate it if there is. Just outline/plan of the code is sufficient. Thank you.
The challenge with what you're trying to do is that the indicators and signals are presumed to be vectorized functions. So they are applied to the entire dataset at once. You're talking about subsetting the data by months, and applying different parameter values. If your indicators and signals don't have any lookback period, or are constant over the entire test, then you should be able to use a rebalance rule, which will subset the data and let you change things around on your rebalance periods. Assuming that's not the case, your options are somewhat more limited. You have no problems using the same portfolio and account over progressively increasing time periods. You're basically just adding new transactions over time. The slightly challenging point will be in controlling the start and end periods for everything else. You'll need to make sure you don't have signals before the start of the subset period, or after the end of the subset period. A minimal reproducible example would help others help you. Regards, Brian -- Brian G. Peterson http://braverock.com/brian/ Ph: 773-459-4973 IM: bgpbraverock
_______________________________________________ R-SIG-Finance at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-finance -- Subscriber-posting only. If you want to post, subscribe first. -- Also note that this is not the r-help list where general R questions should go.