Boston Meeting: Increasing the IR via Ordinal Optimization

Dan Rie has arranged this meeting, and writes:



Gioel Molinari will be presenting at the May 20th Boston QWAFAFEW. He has worked on investigating the application of the optimization framework developed by Almgren and Chris that results from their extension of the Markowitz efficient frontier predicated on the cardinal mean/variance space to an efficient frontier in a purely ordinal space of preferred/equivalent portfolios. The result is a new methodology for optimization of portfolio weights that relies on information about ordinal relationships (rankings) rather than cardinal predictions of expected returns and variances. This more general and less over specified optimization resolves some of the problems associated with optimization that have plagued practitioners. Molinari (and Almgren and Chriss) demonstrate that this optimization framework delivers substantially higher information ratios to portfolios than traditional optimization when the alpha inputs are ranks.

His results are additional substantiation of the value of the work of Almgren & Chriss on using the extension of Markowitz efficiency into ordinal (rather than cardinal) space. It's the Almgren & Chriss paradigm that makes his work of interest as well as what makes it possible. In combination with the evidence provided by the Almgren & Chriss simulations, Molinari's results provide useful color and extension to the foundations laid by Almgren & Chriss in their papers on rank based portfolio optimization.

This presentation is a worthy QWAFAFEW topic because of the extraordinary importance of the Almgren & Chriss work for anyone who is using ranking information for portfolio construction (i.e. pretty much everyone as far as I can tell). Anyone not using the Almgren & Chriss paradigm is probably missing roughly half of the potential information ratio that could be achieved if they were aware of this theoretical advance.