[R-SIG-Finance] Problems with quantstrat and my own code

[Brian G. Peterson]

I located a subtle bug in the nextIndex function inside applyRules 
(which is in turn called by applyStrategy).  It didn't affect all 
instruments or strategies, so it took a little while to locate and fix it.

R-Forge svn r500 contains the fix.

Thanks for the report.  Attached please find the pdf output from 
chart.Posn() of the 'faber' example code run on instrument 'TYC'

Regards,

   - Brian


-- 
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock


On 12/14/2010 04:57 AM, Brian G. Peterson wrote:
> OK, first, to your question:
>
> There is no book that will teach you modeling in R.  Strategy modeling is
> a very complex and often proprietary task.  There are many books that
> purport to teach strategy modeling.  All the ones I own are pretty bad.
> None of them use R.  Some of them are useful as a compilation of common
> widely known strategy parameters, indicators, signals, and rules.
>
> There's a reason that successful quantitative strategists are well
> compensated: what we do is hard, and involves art, science, math, and a
> little bit of luck.  'quantstrat', and the rest of the toolchain, represent
> tools that you can use to make you more productive as you create
> strategies.
>
> Second, to your problems.
>
> 1>  quantstrat
>
> That's a *Warning*, not an *Error*, and is coming out of xts somewhere,
> probably in subsetting, since we do so much of that.  Because it is a
> Warning, and not an Error, I haven't taken the time to figure out where it
> is and either code around it or suppress it.
>
> 2>  your code
>
> I'm not going to debug your ad-hoc strategy code for you, sorry.  Every
> time I set out to write an ad-hoc backtest of a strategy, I screw
> *something* up. We wrote a framework to concentrate the bugs in one place,
> so that we could eliminate them one at a time, and spend more time actually
> modeling strategies. Another way of putting that would be that we we wrote
> the framework so that we could have a different kind of discussion with the
> user of our open source code.  That discussion would look something like
> this:
>
> "I ran the quantstrat demo('faber') code on instrument 'TYC' and
> instrument 'XLY', pdf's of chart.Posn() attached.
> I believe the demo code for market orders is creating incorrect results
> here<insert detailed description>  by<insert your hypothesis of what is
> wrong>. Could someone please check this and let me know?  I'd be happy to
> help patch the code.  Thanks."
>
> Since I think that's probably more or less what you meant, I'll take a
> look at the faber demo on TYC and XLY when I get to the office, and look
> for any obvious problems with the trades and P&L.  If you've found
> something specific that might raise concerns (all the data is available
> either  in the output of applyStrategy or in the portfolio to double-check
> the indicators, signals, and rules), please share.
>
> Regards,
>
>    - Brian
>
>
> On Tue, 14 Dec 2010 00:53:33 -0600, Aaditya Nanduri
> <aaditya.nanduri at gmail.com>  wrote:
>> Hello All,
>>
>> I have 2 problems and 1 question today.
>>
>> QUESTION : Are there any book that teach modelling in R? Hopefully an
>> introductory book. I would like to learn how to build/backtest
> strategies,
>> so if there are any introductory books, please share.
>>
>> PROBLEMS:
>> 1) Quantstrat
>>
>> I tried to use the Faber demo but changed all the symbols to just "TYC"
>> and I get the following errors:
>>     .
>>     .
>>     .
>>   # Process the indicators and generate trades
>>> start_t<-Sys.time()
>>> out<-try(applyStrategy(strategy=stratFaber , portfolios='faber'))
>> *Warning message:*
>> *In max(i) : no non-missing arguments to max; returning -Inf*
>>> end_t<-Sys.time()
>>     .
>>     .
>>     .
>> *Warning messages:*
>> *1: In max(i) : no non-missing arguments to max; returning -Inf*
>> *2: In max(i) : no non-missing arguments to max; returning -Inf*
>>
>> I have no idea what they mean...
>>
>> 2) My code:
>>
>>
>> I know this code is wrong but I dont know where the error is. To me, it
>> seems right. But when I ran "XLY" through the faber demo and the code I
>> wrote, the end results couldnt be further from each other.
>> I also know that my code is highly inefficient so if there are any
> helpful
>> hints that you would like to throw my way, I am open to everything.
>>
>>
> ------------------------------------------------CODE-----------------------------------
>> library(quantmod)
>>
>> # Get Data
>> getSymbols("TYC", from = "1998-01-01", to = Sys.Date())
>>
>> # Convert from daily to monthly
>> x<- get(TYC)
>> x<- to.monthly(XLY, indexAt = 'lastof', drop.time= TRUE)
>> indexFormat(x)<- '%Y-%m-%d'
>> colnames(x)<- gsub("x", "XLY", colnames(x))
>> assign("XLY", x)
>>
>> price<- as.numeric(Cl(XLY))
>> # Build indicators
>> sma10<- SMA(price, n=10)
>>
>> for(i in 1:length(price)){ buy[i]<- NA; sell[i]<- NA }
>> # Buy signal
>> buy[which(price>  sma10)]<- 1
>> buy[which(is.na(buy))]<- 0
>>
>> # Sell signal
>> sell[which(price<  sma10)]<- -1
>> sell[which(is.na(sell))]<- 0
>>
>> signal<- buy + sell
>> signal[which(is.na(signal))]<- 0
>>
>> x<- NA
>>
>> for(i in 1:length(signal)){
>>   # Start with buying not selling
>> if(signal[i-1] == 0&&  signal[i] == -1) {
>> signal[i] = 0
>> }
>> }
>>
>> for(i in 1:length(signal)){
>>
>> if(signal[i-1] == 0&&  signal[i] == 1) {
>> x[i]<- 1
>> }
>> if(signal[i-1] == -1&&  signal[i] == 1) {
>> x[i]<- 1
>> }
>> if(signal[i-1] == 1&&  signal[i] == -1) {
>> x[i]<- -1
>> }
>> }
>>
>> x[which(is.na(x))]<- 0
>>
>> money<- 100000 ->  initEq
>> stock<- 0
>> numinP<- 0
>>
>> for(i in 1:length(x)){
>>   if(x[i] == 1) {
>> p<- price[i]
>> b<- floor(money/p)
>> numinP<- numinP + b
>> money<- money - p*b
>> stock<- p*b
>>   }
>> if(x[i] == -1) {
>> p<- price[i]
>> money<- money + p*numinP
>> stock<- 0
>> numinP<- 0
>> }
>> }
>>
>> total<- money + stock
>> print(paste("The initial equity:",initEq))
>> print(paste("After 10 years of trading using TAA:",total))
>>
>>
>> I appreciate all your help
>>
>> --
>> Aaditya Nanduri
>> aaditya.nanduri at gmail.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: quantstratTYCfaber_test.pdf
Type: application/pdf
Size: 42870 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20101215/fb300286/attachment.pdf>