Hello Horace,
It is only missing by omission; can't think of everything in a short
post. I typically use market data or bootstrapped data for that, but
you approach is equally, though differently, useful. I've been
thinking a lot about Ralph Vince's comment on the Markets list about
the arc sine law. I have observed that phenomenon in my bootstrapped
series and I think that it is a very deep well.
Best,
John -- Who is a bit worried that these comments are not Rsih
enough for this environment. Please forgive my gaucheness, but this is
a topic near to my core.
On Thu, Sep 29, 2011 at 4:35 PM, Horace Tso <Horace.Tso at pgn.com> wrote:
John, always appreciate the insights from an expert in the trade, but what seems to be missing from your list is a simple series generated from random walk, e.g. x_t = cumsum(rnorm(10000,mean=0.05)/100) That's pretty much the null hypothesis for testing any trading system. Horace -----Original Message----- From: r-sig-finance-bounces at r-project.org [mailto:r-sig-finance-bounces at r-project.org] On Behalf Of BBands Sent: Thursday, September 29, 2011 3:00 PM To: R-sig-finance Subject: Re: [R-SIG-Finance] Test data On Tue, Sep 27, 2011 at 2:14 PM, Worik Stanton <worik.stanton at gmail.com> wrote:
I am about to generate some data to test some technical analysis functions. I expect I am not the first! ?Has anybody some advice about where to look for some data sets? What I need, naturally, ?is pairs of series, input and output. ?I expect I can roll my own without too much difficulty but...
In developing indicators and technical methods it is very helpful to feed models various synthetic data streams. I have a library of these and will discuss a few that I use here. 1. Instantaneous change. The is a dead simple series, but it can be amazingly informative. Mine runs at 1 and then jumps to 2. You might be surprised at the results you get, especially if any form of advanced smoothing is used. 2. Sine, triangle, saw tooth and square waves of various periodicities. 3. Idealized typical technical patterns, head and shoulders, double bottoms and tops, wedges, simple reversals and the like. 4. Cyclically varying volatility. 5. Most of the above coupled with trends, say ?plus or minus 10% annualized growth rates. The ideal is of course to torture test transforms so you will know what to expect from them in evolving market conditions. This is especially helpful in avoiding signals that are artifacts of your transforms rather than of the data. Try plotting 20 period, 2 standard deviation Bollinger Bands with 10, 20, 30 and 40 period square waves. Best, ? ?John
_______________________________________________ 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.
John Bollinger, CFA, CMT www.BollingerBands.com If you advance far enough, you arrive at the beginning.