Portfolios— Optimal in Real Time


Studying existing portfolio management schemes and researching and developing new ones, using robust and fully-disclosed methods of quantitative analysis— on behalf of professional money managers and investment advisors who are disposed toward asset allocations that are optimal yet responsive to shifting market conditions… who either want their own systems tuned and qualified and made more reliable or want new ones created for them.

Card image cap

Mike O'Connor

Improving quantitative financial analysis and the coding for it.

MO'C Portfolio Analytics is a service of a Washington State, USA sole proprietorship.

Michael C. O'Connor, Proprietor

biography  ➣➣ research  ➣➣ innovation  ➣➣ production  ➣➣ languages  ➣➣ codebases  ➣➣ ongoing 



The Outlook

Financial analyses are all about history mattering. Even determined fundamentalists must be willing to admit that the principles under which they operate should have worked well in the past. And the history under study need not be limited to data on returns but could include security-specific fundamentals or factors that influence the market as a whole. But here's the catch: If you don't know what you're doing it's fairly easy to fool yourself— to stumble upon a scheme that by chance alone would have worked well in the past but won't in the future. That, we call a “false discovery.” The first order of business is therefore to develop and deploy mathematical means of avoiding false discoveries.

Avoiding False Discoveries

There are “in-sample” and “out-of-sample” methods of controlling the risk of fooling yourself with backtesting. Harvey and Liu had suggested, in this 2014 article, that an out-of-sample method of cross-validation could perhaps be combined with a certain in-sample method of controlling the false discovery rate: “A successful merger of these two approaches could potentially yield more powerful tools to help asset managers successfully evaluate trading strategies.” My new article accomplishes exactly that. With it, investment managers can very substantially limit their chances of making and implementing false discoveries and in so doing form portfolios that perform better than the market.

Avoiding Missed Opportunities

“Hypothesis testing” has a well-formulated meaning in statistics. But to test to see if an optimized portfolio is likely to continue to outperform or not is tricker than many suppose. I recently worked on exposing certain limitations to effective hypothesis testing that we encounter when using the “Sharpe ratio”, which is a risk-adjusted return performance measure. In this article of mine I show that unless the returns of the two portfolios are highly correlated the test gives the wrong answer all too frequently: Too often it wrongfully grades a genuine and replicable improvement in the Sharpe ratio as being statistically insignificant.

Meeting Challenges

During the last two steep stock market declines of 2000–03 and 2007–09 diversification didn't help; only getting into cash or, say, treasuries would have worked. Can that be done in real time? A working paper of mine, accessible here, reports on my early research on momentum which met with limited success but did include a way of effectively avoiding debacles. But quite recently I developed a new approach that is utterly definitive with respect to preventing deep drawdowns without inducing long-term adverse effects on returns. The new endeavor is like the earlier one in that it implements a walk-forward procedure, to ensure that estimates of performance are bias-free, but there the similarity stops.

More on the New Method

It is principally based on the use of a repurposed “shrinkage estimator”, where the repurposing has to do with how the procedure is caused not only to shrink the uncertainty in the outcome but also to improve its mean value. The method targets funds such as ETFs or tradable mutual funds, not stocks. Deep drawdowns are avoided while mean returns and Sharpe ratios are improved. It has not yet been determined as to whether or not the basic algorithm will be made public. Certainly variations of it will be produced, customizations, that will not be. Inquiries are welcome.

Serving Clients

Naturally, with a consultancy such as this, quite flexible arrangements that best suit the mutual interests of the client and the Proprietor can readily be realized. Presently the focus is on testing and tuning investment alternatives and developing new ones, on behalf of professional money managers and investment advisors. But in the future a subscription service may be offered to retail investors, involving active investments in exchange-traded funds.