The first international conference dedicated to the use of R in the finance industry, R/Finance 2009, was a great success. With over 150 attendees (my poor estimation skills notwithstanding), sold-out tutorials, and an outstanding lineup of invited and contributing speakers from around the world, this event really demonstrated the importance of R in the world of financial analysis.

Some of the highlights of the event for me were:

Robert Grossman kicked off the event with style, declaring that "R is mainstream!". Not only are unrelated companies promoting R, but R is now a major application of the fastest-growing meme of 2009: cloud computing. Robert has an excellent demonstration of using R and Hadoop on Amazon's EC2 cloud computing service (although due to technical difficulties couldn't show it live). 

Diethelm Wuertz gave an amazing talk about portfolio optimization in R with Rmetrics. If you haven't taken a look at Rmetrics before, it is the premier open-source software for financial market analysis and financial instrument valuation, and includes over 32 R packages comprising hundreds of functions. I was truly blown away with the capabilities of Rmetrics for portfolio analysis, and it seems to me that with the capabilities Wuertz described, R now represents the state of the art in portfolio optimization technology. (And given that he uses these tools to manage his own fund, they must be really good!)

Guy Yollin (Rotella Capital Mangement) also spoke about portfolio optimization, and in particular demonstrated the flexibility of R by modifying the standard solve.QP function to introduce custom constraints. This flexibility is what makes R so powerful for financial analysis, and Guy showed an example of optimizing a portfolio where all weights must either be zero, or greater than 3%. He also showed applications in Conditional Value at Risk (CVaR), and maximum drawdown optimization. 

My colleague Bryan Lewis (REvolution Computing) introduced the new foreach function from ParallelR 2.0, and turned more than a few heads by using it to dramatically speed up large-scale backtesting of a portfolio trading strategy. He also showed a very neat animation (generated using the spatstat package) 

David Kane and Patrick Burns gave complementary presentations on assessing portfolio performance. David gave a nice overview of the characteristics approach for finding matching portfolios for performance assessment, noting that while it is widely cited in the academic literature it's little-used in practice. As an alternative, he suggests using the MatchIt package (as described in this vignette) to find matching portfolios, and showed how this method outperforms the characteristics method to match a recommended Starmine portfolio. 

In a similar vein (but in his own inimitable style), Patrick Burns focused on generating random portfolios: not just for portfolio assessment, but also for testing trading strategies, evaluating constraints and validating risk models. For assessment, Patrick demonstrated convincingly that random portfolios are superior to using a benchmark or peer groups. To me, an even more compelling application Patrick showed was using random portfolios for evaluating the effect of constraints on the portfolio. Although constraints are supposed to be "insurance against the portfolio doing anything stupid" you may be surprised at what you're paying for this insurance: As Patrick showed using random portfolios, even simple constraints may be sacrificing positive returns without any elimination of lower-tail risks.

Brian Rowe (Merrill Lynch / Bank of America) drew on insights from the world of physics to filter noise in correlation matrices using Random Matrix Theory. His methods are implemented and available in the R package Tawny.

Roger Koenker suggested that quantile regression, currently an under-used technique in finance, has application in the modeling of "bubble-like" phenomena in finance. In addition to allowing parametric models for quantities like Value at Risk defined by quantiles, quantile regression also allows modeling of what he called "pessimistic portfolios", where the probability of "good" events are downplayed while those of "bad" events are amplified. Also: betting on college basketball. The techniques he described are available in the R package quantreg.

Eric Zivot (University of Washington) offered the insights of a recent convert from S-PLUS advising a fund of funds using R for its quantitative analysis. As the author of a book on financial modeling with S-PLUS, it was interesting to hear him say that overall R now has more financial functionality than S-PLUS (while still lacking a few functions that S-PLUS has). It was also interesting to hear his experiences working with an IT department: their reluctance to support R or let it talk directly to the data warehouse (meaning that all data I/O was via Excel).

Wow, that's a lot of highlights for a single conference! If you missed the event, I've heard that the presentations from these and the other excellent speakers will soon be available from the R/Finance 2009 website. Kudos and thanks go out to the organizing committee for putting together such an amazing event. I'm very proud that REvolution Computing was a sponsor of this groundbreaking conference, and I look forward to the next one. It's already been confirmed that a second R/Finance conference will be held -- watch this space for further details.