search

timtrice / backtesting-strategies

Book on backtesting strategies in R using blotter, quantstrat, FinancialInstruments, TTR packages
http://timtrice.github.io/backtesting-strategies/
106 stars 61 forks source link

Add Monte Carlo Analysis (currently Ch. 14) #4

Open timtrice opened 8 years ago

timtrice commented 8 years ago

Need to add content, example for performing a Monte Carlo analysis.

braverock commented 8 years ago

blotter now contains the 'mcsim' function along with plot, hist, and quantile methods, and we're actively adding more material.

https://github.com/braverock/blotter/blob/master/R/mcsim.R

monte carlo simulation in blotter now supports simulation from daily portfolio results using use=c('equity','txns','returns') for the input daily P&L, sampling with/without replacement, sampling with/without block samples, with/without variable block length. it returns an S3 object that now has plot, hist, and quantile methods

We will work on random trade simulation in the near future as well, and more functions for detecting or discounting possible overfitting.

jaymon0703 commented 6 years ago

Hi @timtrice looks like a really awesome book you got going here. I am keen to update some of the missing information as relates to MC analysis in particular. Also, in the quantstrat README, i would like to add it as a reference if you approve?

Lastly, may be useful to add an Overfitting chapter...Brian and I have been doing some work on that in the last year or so.

timtrice commented 6 years ago

@jaymon0703 , apologies for not responding sooner. The book is in a drastic need of an update; at one point not too long ago I had begun working on a redo but had to set it aside.

You're welcome to contribute however you feel adequate. I think a chapter on overfitting would definitely be of value.

Most of the original content will remain but with updates correspondingly to package updates. The tidyquant packages I think need to be added as well. So, there's quite a bit.

Whatever you have to contribute, I would definitely appreciate it!

braverock commented 6 years ago

@timtrice : what do you think tidyquant adds besides latency and dumbed-down interfaces for dealing with time series? As far as I know, there are zero new analytical methods in tidyquant, all the actual content is ported from other packages, or depends on those packages? A 'wide' representation for time series data doesn't make a lot of sense, and results in multiple unnecessary copies and allocations of memory, so I'm wondering what value you see in it?

jaymon0703 commented 6 years ago

@timtrice - thanks again... @braverock has inspired much of the Monte Carlo and Overfitting work, and discusses it in his latest seminar at R/Finance 2018 available here, so i will see what we/I can put together over the coming weeks and submit pull request/s accordingly.