[R-SIG-Finance] Generation of synthetic financial data

[Andrew Piskorski]

On Mon, Mar 28, 2022 at 12:43:45PM +0200, Andri Schnider wrote:

> I was curious what you consider to be the most useful tools for generating
> synthetic time series that behave "similarly" to an original series that
> one has (one difficulty is certainly choosing metrics that define
> "similarity" between two series, but that is an entirely different
> discussion).

I don't have a proper answer to your question.  However, as an aside,
you might find these two old books of interest:

  https://www.amazon.com/Economic-Function-Futures-Markets/dp/0521389348/
  by Jeffrey C. Williams; (c) 1986

  https://www.amazon.com/Storage-Commodity-Markets-Jeffrey-Williams/dp/0521023394/
  by Jeffrey C. Williams, Brian D. Wright; (c) 1991

That second book is about commodity "storage model" theory, and walks
through a (totally deterministic, no random numbers involved) algorithm
for generating time-series with qualitative behavior similar to observed
real-world commodity prices.  It also talks about how ARMA and ARCH
models capture some of the statistical properties of the generated
series, but miss other parts.

So if you want an unusual, theory-motivated way of generating
realistic-looking price series with certain statistical properties,
that's one place to look.  (No, I haven't done it.  But their
pseudo-code starting on page 81 looks detailed enough to turn into a
real implementation if you wanted to.)

-- 
Andrew Piskorski <atp using piskorski.com>