27 April 2009

'R' you experienced?

" ... To confer, converse, and otherwise hob-nob with my brother wizards ..."

I spent a productive weekend up in Chicago, at the R/Finance 2009: Applied Finance with R conference, which billed itself as the "first annual R/Finance conference for applied finance using R". The conference organizers and hosts are the 'usual suspects' on the 'R sig fin' mailing list; Jeffrey Ryan, Dirk Eddelbuettel, Dale Rosenthal, Brian Peterson, Peter Carl, Gib Bassett, and John Miller, assisted by the talented and imperturbable Holly Griffin of UIC. Pretty clearly most of this group code together regularly; see the committer list on the blotter module. The venue was at the 'other school' in Chicago, the one with a more practical interest in Economics and Finance

An aside about Chicago: Long ago, and far away, I was trained as a acolyte 'monetarist' by disciples of Herb Stein's CEA and the Fed, in the [University of] Chicago school [a fad, seemingly long forgotten by recent Economics and Finance grads, so far as I can tell]. Monetarism is a forgotten religion these days; the Fed stopped formally publishing its M3 series a few years ago, in light of the rise of what Bill Gross calls the 'shadow banking' system.

Luke: [The robot] claims to be the property of an Obi-Wan Kenobi. Is he a relative of yours? Do you know what he's talking about?
Obi-Wan: Obi-Wan Kenobi. Obi-Wan... Now, that's a name I've not heard in a long time. A long time.
Luke: I think my uncle knows him. He said he was dead.
Obi-Wan: Oh, he's not dead... Not yet.
Luke: You know him?
Obi-Wan: But of course I know him

Many of the organizers were known to me from my email correspondence or from observing their packages, and I had spoken with one (Dirk) two or three years ago briefly after the trading shim was first usable

Dirk seems to be a bundle of unbounded energy. His tools had solved a lot of data storage and visualization issues for me early on in our project. He led a push a couple years back to drill in many of the R add-on modules into the main Debian archives. I still hope to emulate his example in rpm space using R2spec and some post-processing scripts (Dependency enumeration is not quite perfect yet). Dirk has already mentioned in his blog the 'after-sessions' at Jak's; we closed the place down Saturday with useful brainstorming happening long into the night

I had resolved to travel to the conference to learn, and to stay quiet as to matters of FOSS and advocacy. I was almost even able to keep to that intent, save at the interstitial times. The formal presentations were amazing in their quality of content, competence of the presenters, and challenging to my old knowledge of Statistics and Mathematics. I even understood most of what the presenters were doing, and why, on the formal finance side and will re-read the slide decks with great interest when they appear to fill in the holes. One would have probably had to be there to draw much more from the decks, as the presenters were not doing the occasional 'stand and read' presentation one finds at some conferences, but rather largely used their decks as reminders of the points they wanted to hit and elaborate on in their presentation, and to state exactly the code and formulae in play. The committee have really set a high bar to reach for next year's event to top, and I look forward to it already

The pre-conference tutorials were worthwhile. I knew Jeff Ryan's work from xts and IBrokers of course, and gained insight into his mental roadmap on where the code is going and how it will get there. I think the enhancements he is trialling in xts will pretty clearly flow back upstream into zoo in general form; I had not heard of Dale before, but his breakout and presentation of an analytic approach on addition and testing of single constraints (I have covered scientific method and epistemology here before, and will again) served as a fine warm-up to the formal sessions

During the session breaks, at meals, and into the night, I had a chance for give and take at length with several of the committee, presenters, and attendees, to bridge what Patrick Burns spoke on -- the chasm between Practice and Theory

Part of the trip and my need for listen, was to get a handle on how to match the shim and R as a pair of heavy-weight co-processes, so that the user of the shim can hook in and use the wonderful tools already in R space. We'll most likely get there, but the timing is not clear. Having said 'we', permit me to make it clear that the heavy lifting will be done, if and when done, by Bill, and not me. At the prodding of Peter, who as I understand it regularly team codes with Brian, I have started the sign-up process for an account at r-forge, and will 'cut my teeth' on a simple connector or module, to warm up my skills as a co-development tester and 'guinea pig' consumer of the major task of integrating the shim

FIX rules the roost for being the 'lingua Franca' for interchange to exchange order, position and fill data with counter-party upstream brokers or exchanges (thanks here to the CME Foundation for partially funding the event). We will not soon be adding a compressed FIX connector to the shim, and certainly not before we attain our major milestone of a formal 'complete' first release.

Finally, a couple folks asked why we were playing down in retail space with the TWS and its vendor specific API. For a researcher, and for a small proprietary trader, we still find IB's API and services the most affordable, and substantially complete. It is a gateway to enable any interested researcher to do material research (the 'Theory') and strategy development and execution (the 'Practice'). For student academics, the availability of IB's 'trading Olympiad' program and the shim, and R offer all one needs for a better than free price

Update: We see also the summary of the event at Revolution Computing, an R vendor