Comments on FOSS Trading: Tactical asset allocation using blotter

Thanks a lot for the clarifications Brian!

2011-11-10T22:33:59.207-06:00

Thanks a lot for the clarifications Brian!

The script itself says: "NOTE: For the purp...

2011-11-08T21:00:59.481-06:00

The script itself says:

"NOTE: For the purposes of this demo, we ignore interest and leverage."

It doe *not* say "for the purposes of this demo, we use no leverage".

In basically all institutional accounts and in many retail accounts, open equity is essentially equivalent to cash, and it would be fine to use the ending equity, including any open equity, to decide on the sizing of any new positions.

In any event, patches are always welcome to extract available cash and use that instead (ending equity minus position values in a long-only portfolio, as I recall).

Joshua, thanks a lot for your reply! My results ar...

2011-11-08T19:33:22.468-06:00

Joshua, thanks a lot for your reply! My results are 100% reproducible to me. The code is similar to yours, I've just added the following 3 lines at the top, before loading the libraries, to make the code independent of the workspace:
rm(list=ls(all=TRUE))
if(!exists('.instrument')) .instrument <<- new.env(hash=TRUE)
if(!exists('.blotter')) .blotter <<- new.env(hash=TRUE)
As verbose = TRUE I can see the results of the strategy allocations in the console output, I've just ran it again and got the following:
... SKIP ...
+ } # End dates loop
[1] "2003-04-30 IEF 578 @ 86.39"
[1] "2003-04-30 SPY 544 @ 91.91"
[1] "2003-07-31 IEF -578 @ 83.38"
[1] "2003-09-30 IEF 597 @ 86.59"
[1] "2003-10-31 IEF -597 @ 84.6"
[1] "2004-02-27 IEF 627 @ 86.68"
[1] "2004-04-30 IEF -627 @ 83.5"
... SKIP ...
The generated PNG images are absolutely the same as in your post.
If I understand correctly:
1. we've bought 578 shares of IEF on 2003-04-30 at $86.39 that leaves $50066.58 at our account unused.
2. than we've bot 544 shares of SPY at $91.91 that leaves us with $67.54 on account
3. on 2003—07-31 we've closed our IEF position with loss at $83.38 that leaves $48261.18 at our account
4. next, we are buying 597 shares of IEF at $86.59 that leaves us with minus $3433.05 etc.

Maybe I'm misunderstanding something, please clarify. Thanks!

seega, I don't where you're getting those...

2011-11-01T20:45:13.735-05:00

seega,

I don't where you're getting those numbers. I just ran the script myself and cannot replicate what you've shown in your comment.

Thanks for the nice example, but there's diffe...

2011-10-31T20:49:23.914-05:00

Thanks for the nice example, but there's differently an error in the calculations of return for tactical strategy, as an example here are several lines of results that I've got from the strategy:
Date Symbol Size Price Change Total
0 0 0 0 0 100000
4/30/2003 IEF 578 86.39 49933.42 50066.58
4/30/2003 SPY 544 91.91 49999.04 67.54
7/31/2003 IEF -578 83.38 -48193.6 48261.18
9/30/2003 IEF 597 86.59 51694.23 -3433.05
10/31/2003 IEF -597 84.6 -50506.2 47073.15

When second position opened in IEF on 9/30/2003 there's -3422.05 left at the account that suppose to never go below 0. I understand that we can borrow money on margin, but this is not the case for the strategy, and I see several instances in the results when buy made on margin. Probably because of that at the end of the last term the return if calculated manually is more like 17%, and not 22.7% as per the script result.
Sorry, maybe I'm missing something, please point it if so. FYI, I'm using R version 2.13.2
Thanks again for the great job!

Hasan and Elihu, Sorry for the extremely slow rep...

2011-09-25T12:28:55.475-05:00

Hasan and Elihu,

Sorry for the extremely slow reply. I've corrected the code to work with more recent versions of blotter. Also see the update at the top of the post.

Best,
Josh

Second on Hasan's error - the whole script run...

2011-08-31T00:04:00.295-05:00

Second on Hasan's error - the whole script runs up until:

"charts.PerformanceSummary(ROC(getAccount('default')$TOTAL$End.Eq)[-1],main="Tactical Asset Allocation")

Then R throws "Error in dim(data) <- dim : attempt to set an attribute on NULL."

I'm running R 2.13.1 (installed today) in RStudio, and I just ran the update.packages() script mentioned above.

Thanks Josh for the great tutorial!

Ok, I can get the returns and see the actual data ...

2011-04-15T17:37:15.162-05:00

Ok, I can get the returns and see the actual data structures. But the code crashes at the first chartSeries line. So, I tried looking at getAccount('default'), which looks like:

# from head(getAccount('default'))
$portfolios
$portfolios$default
Long.Value Short.Value Net.Value Gross.Value Realized.PL
2002-07-31 0.0 0 0.0 0.0 0.00
2003-04-30 199951.3 0 199951.3 199951.3 0.00
2003-05-30 208813.3 0 208813.3 208813.3 0.00
2003-06-30 208558.1 0 208558.1 208558.1 0.00
2003-07-31 108136.3 0 108136.3 108136.3 -3482.57
2003-08-29 110366.7 0 110366.7 110366.7 0.00
2003-09-30 212134.1 0 212134.1 212134.1 0.00
2003-10-31 114566.4 0 114566.4 114566.4 -2376.06
2003-11-28 115817.6 0 115817.6 115817.6 0.00
2003-12-31 121072.6 0 121072.6 121072.6 0.00
2004-01-30 123466.2 0 123466.2 123466.2 0.00
2004-02-27 233925.2 0 233925.2 233925.2 0.00
2004-03-31 232928.0 0 232928.0 232928.0 0.00
2004-04-30 120724.5 0 120724.5 120724.5 -3990.90
2004-05-28 122791.7 0 122791.7 122791.7 0.00
2004-06-30 124608.6 0 124608.6 124608.6 0.00
2004-07-30 0.0 0 0.0 0.0 20595.84
2004-08-31 105362.5 0 105362.5 105362.5 0.00
2004-09-30 105374.8 0 105374.8 105374.8 0.00
2004-10-29 211374.7 0 211374.7 211374.7 0.00

Works fine on my linux slice but I noticed install...

2010-11-06T18:25:28.032-05:00

Works fine on my linux slice but I noticed install blotter is blowing up on my mac. - https://r-forge.r-project.org/R/?group_id=316&log=build_mac&pkg=FinancialInstrument&flavor=patched

Just wanted to throw out a heads up.

SWAustinGuy, I've spoken with Brian and we be...

2010-10-11T11:29:12.396-05:00

SWAustinGuy,

I've spoken with Brian and we believe the error is in r411 of blotter and has been fixed in r412. But you didn't get the fix because the Windows binary of r412 has not been built on R-forge yet. Please try again once you see "Rev: 412" for blotter on this page.

Sorry for the trouble.

Hmm, if I change the following line in updatePortf...

2010-10-10T19:51:46.854-05:00

Hmm, if I change the following line in updatePortf (Dates=..., the code runs much further, executing a few hundred trades, then failing.

old:
updatePortf(ltportfolio, Dates = CurrentDate)

new:
updatePortf(Portfolio='default', Dates=paste('::',as.Date(Sys.time()),sep=''))

Then fails with:
[1] "2007-05-31 IEF -611 @ 91.45"
Error in `[.xts`(sym, i, ) : subscript out of bounds

Is the date thing an issue in Windows?

Hi - I could use some help in getting the example ...

2010-10-10T18:56:54.137-05:00

Hi -
I could use some help in getting the example 'longtrend' working in R-2.11.1, and the latest blotter, xts, quantmod, from R-forge (just updated all packages today).

I'm running Win7 (but don't have to if it's an issue). The problem I am having is in updatePortf()
I get the following error

[1] "2001-04-30 IEF 578 @ 86.39"
Symbol IEF Date: 988588800 TxnPrice 86.39 Qty 578
Here

Error in if (ncol(Prices) > 1) Prices = getPrice(prices, Symbol)[dateRange] :
argument is of length zero

It appears as if everything is working correctly except for the update to the portfolio. I've dumped my enironment, version info, code I am running (from example), error message, data frame for underlying symbol IEF, and the error I get when trying to manually add a transaction and update portfolio on the command line of 'R'.

http://oeo.la/T2zHyC

I'm very new to 'R', so I hope to have captured all relevant information and thank anyone in advance that might be able to steer me in the right direction.

Great blog - Thanks allot for sharing this.

Another alternative for cleaning up the environmen...

2010-08-16T19:25:07.661-05:00

Another alternative for cleaning up the environment would be:
rm(list=ls(all=TRUE, pos=.blotter), pos=.blotter)
rm(list=ls(all=TRUE, pos=.instrument), pos=.instrument)

Zachary, I have the same problem. I can run the s...

2010-07-31T15:33:02.950-05:00

Zachary,
I have the same problem. I can run the script once, but it fails on subsequent attempts because the portfolio already exists. One way I found to remedy the issue is to cleanup the workspace.

# Reset the workspace by deleting all variables and recreating
# the environment variables created during library load
rm(list=ls(all=TRUE))
.blotter <- new.env()
.instrument <- new.env()

-Ben
benmccann.com

Thanks Josh, I got it to work. There was an odd m...

2010-06-10T12:01:52.327-05:00

Thanks Josh, I got it to work. There was an odd masking going on I was able to clear out. It works beautifully.

G$, quantmod:::getPrice is in the most recent Mac...

2010-06-04T13:21:19.077-05:00

G$,

quantmod:::getPrice is in the most recent MacOS X binary of quantmod on R-forge (I just looked). I have no idea what you could be doing wrong.

getPrice isn't exported, which is why you can&...

2010-05-10T08:10:47.726-05:00

getPrice isn't exported, which is why you can't find it via the methods you tried. It's in the source file Price.transformations.R (functions do not need to be in a file of the same name).

quantmod:::getPrice will show you the source code from the R console.

Sorry, I did download the latest off r-forge, vers...

2010-05-09T23:17:37.872-05:00

Sorry, I did download the latest off r-forge, version 03.03-14.

I loaded the package, library(quantmod)

I can't find any getPrice.R in the R directory, neither can I use help(getPrice) to find any documentation for it to pull from the man directory. I do see a getQuote?

G$, getPrice is in quantmod. Run update.packages...

2010-05-09T22:03:42.652-05:00

G$,

getPrice is in quantmod. Run update.packages(). If that doesn't work, get the most recent version of quantmod from R-forge.

Josh, I just downloaded the latest MacOS X Leopard...

2010-05-09T22:00:17.156-05:00

Josh, I just downloaded the latest MacOS X Leopard binary for blotter and am trying to follow the example so I can use the package with my Fixed Income Calendar Effects strategy, but I am getting some environment errors such as:

Error in updatePosPL(Portfolio = pname, Symbol = as.character(symbol), :
could not find function "getPrice"

Am I missing an installed package, or do I have to do something to expose the getPrice function?

Trey, Yes, that will work until I can fix the adj...

2010-04-29T10:26:36.743-05:00

Trey,

Yes, that will work until I can fix the adjustment algorithm. Accuracy is only an issue for securities with many splits and/or dividends over an extended period of time.

You're correct that code couldn't be executed in real-time. That said, this strategy trades very infrequently and you would generally know well ahead of time whether the month-end close would be above/below the SMA, so you could enter the trades at the month-end close. Your observation is very important at higher frequencies, however.

The R/Finance 2010 presentations should be on the website in a few weeks. I'll post when they're available.

Josh, I changed the code to the following: IEF =...

2010-04-29T10:12:15.823-05:00

Josh,
I changed the code to the following: IEF = adjustOHLC(IEF, use.Adjusted=TRUE) and I did the same for SPY. Everything worked fine. The documentation does state that this method is less accurate.

As is, this system isn't tradable. I'm not sure if you intended that or not. In my opinion, it's important to make sure that a back-test is actually tradable, in the sense that it could actually be executed in real time. The rule is "If Close > SMA buy on close". The values are determined after the market has closed which requires you buy at the next bar, not that bar close. In this example, that would be the open of the following month. It seems that we would need to switch to daily for the entry. I'll see what I can do. Thanks again for the code and I hope this helps.

Are the presentations from R/Fin 2010 going to posted with prior years?

Trey, Thanks for the information. It looks like ...

2010-04-13T21:57:41.262-05:00

Trey,

Thanks for the information. It looks like something changed in merge.xts that is affecting the adjustment algorithm. I'm busy preparing for, and attending, R/Finance 2010, so I won't be able to dig into this further until next week. Thanks for the report.

Best,
Josh

Josh, Thanks for posting the example. I've be...

2010-04-13T20:32:45.967-05:00

Josh,
Thanks for posting the example. I've been working with this example and ran into the same error as judsonm123. It appears that IEF = adjustOHLC(IEF) results in NAs for the OHLC columns which then creates an error in to.monthly and SMA calculations. I'm running R 2.10.1 with the latest required libraries. I'm still trying to figure it out but I thought I'd let you know.

Best Regards,
Trey

Nim, Sorry, it was a bug in the code caused by so...

2010-03-07T15:44:24.778-06:00

Nim,

Sorry, it was a bug in the code caused by some recent updates to blotter on R-forge. The example works now.