Sigma1 Financial Solutions Software

Seven months ago I posted a blog responding to a question asked by almost every VC or angel investor who just heard my “elevator pitch.”  They asked “Will this software make money for its users?”.

I hate to answer just a point-blank question with a simple yes or no answer.  However, in a 10-minute-or-less Q&A session with angels/VCs, the only acceptable answer I’ve found to-date is an unqualified “Yes.”  This answer is almost inevitably followed up the the question “How much more money?”  (The savviest put the more in the question).  To which I answer “on a risk-adjusted basis about 1-2% more per year.”

This is where different conversations often diverge.  So far about 60% of the time the conversation leads to the angle/VC telling me investors don’t care about risk, “they only care about return — how much return, how soon?“ At this point my private gut reaction is, well there’s another angel that just doesn’t get it and probably won’t in the next ten minutes.  I do my best to explain that future risk is just as important as future return. [From a epistemological perspective there are no "future returns", only future expected returns.]  With the 60% crowd, I usually get to point where I utter the cliche, “Will you acknowledge that there are times where return of capital trumps return on capital?”

Then there is the 40% crowd.  They don’t argue the “risk-adjusted return” point. These angel investors acknowledge that risk management is an important component of intelligent investing. Their next question is typically, “So what sets Sigma1 software apart from other portfolio-optimization products?”  Here’s the response I have delivered so often that I probably recite it in my sleep:

Sigma1 HAL0 software offers two key differentiators. One: HAL0, unlike competing products, was designed from the ground up to support almost any risk measure a user can devise.  Two: HAL0 Portfolio-Optimization Software supports concurrent three-objective optimization, whereas the vast majority of competing optimizers are limited to two.  Those that “support” multiple, alternate risk optimization typically do so by cheating — by optimizing risk/return curves pair-wise, rather than concurrently.

For every VC/Angel conversation that has reached this point, the Q&A has always been different.  However the feedback I get has some commonality.  First, potential investors like the wide-open risk measurement and management capability.  Second, they are luke-warm about the three-objective optimization capability.  They ask, “Why not just let the user choose the best risk measure and optimize to that?”.

It would be all to easy to drop the 3-objective capability from both the investor pitch and the software. Doing so would make user-interface development easier because I’d only have to support the 2-D interface (rather than supporting both 2-D and 3-D interfaces).

Probably the main reason I stubbornly retain the 3-objective concurrent optimization capability is because I use it.  I use it because I have yet to find a single-best risk metric. Most Sigma1 beta partners to-date favor standard mean-return variance. Personally I favor both modified semivariance and max drawdown as metrics.  I am also exploring a mix of monthly and ttm (twelve-trailing month) semivariance as individual concurrent optimization metrics. I find it extremely helpful to analyze the risk A versus risk B trade offs.  I am interested in risk measures that are relatively uncorrelated and produce substantially different portfolios for the same expected returns.

My hypothesis is that portfolios that are optimized for sufficiently-high expected return and sufficiently low expected risks to two diverse risk metrics are more likely to be more robust throughout a variety of market conditions.

Have you ever read a fund prospectus? [Rhetorical question]  Did it list just a single risk?  – Heck No!  It probably read like a shopping list including items such as: market risk, adviser risk, geo-political risk, interest-rate risk, currency risk, illiquidity risk. I could go on, but I hope you see my point.  As investments are subject to a wide variety of risks, why should portfolio-optimization software seek to reduce only one type of risk (typically volatility, aka annualized standard deviation of total return)?

Let me get back to the first question, “Will portfolio-optimization software make me more money?”.  Let me refer to my post from Sept. 22, 2012.  It listed three sample portfolios generated by HAL0 Portfolio-Optimization Software on May 14th, 2012.  It was generated from a list of securities that I considered interesting and moderately complementary.  Expected return for each security was forecast with proprietary formula I created.  (While I’m keeping the expected return formula secret, I will divulge that square-root of beta, and ttm P/E are the two dominant inputs.)

The optimization used both 36-month, monthly, total-return variance and (modified) semivariance (MSV) as risk co-metrics. It produced dozens of optimized portfolios, however I selected 3 based solely on their efficient frontier of return to modified semivariance.  The “highest-risk” portfolio was the portfolio with the highest expected return, the “lowest-risk” portfolio had the lowest MSV, and the “medium-risk” portfolio was the portfolio on the MSV / Expected-Return efficient frontier with the median MSV.

DISCLAIMER:  This data set is by no means sufficient to make any statistically-significant claims.  Conversely, it is not cherry-picked.  The initial data was posed Sept. 22, 2012, and is to the best of my knowledge, the only publicly-reported portfolio specific data from HAL0 optimization.

I want to highlight a few items.  First, all three portfolios beat the total return of the S&P 500 Index.  Second, they performed in an intuitive way for an upward-market period, with the riskiest portfolio ranking highest, medium in the middle, and least-risk ranking lowest of the three.

The highest risk portfolio composition is, obvious, terribly undiversified.  The “medium-risk” portfolio is also higher-risk than would be advisable for most investors.  The lowest-risk portfolio, however, is not an unreasonable portfolio composition for investors comfortable with speculative 100%-stock-based portfolios.  The 4.6% excess return over the S&P 500 Total Return benchmark would please most clients and investors.

In homage to the 60% mentioned above, I haven’t computed the Ex-post MSVs of the portfolios or of the benchmark.

I wrote this post because I wanted to address the common questions of VC and angel investors.  One such question is “Where’s your data… not back-tested data… but documented, real predictive data?”

Almost all of the data I have generated and shared with third parties has been in the context of beta partnerships.  For obvious confidentiality reasons, I cannot share the data, not even the list of securities being optimized. Perhaps, one day, I will ask for and receive permission from one the Sigma1 beta partners.  I’m sure that would require lots of time going back and forth with lawyers.  I’d much much rather spend my time working with investment advisers, portfolio managers, and other investment pros.

Chart 1.  Ex-post Returns through 4/19/2013

Chart 2.  MSV Risk Metric in second column  (modified return semivariance, Ex-ante).

None of the above, nor is any other content on this blog considered investment advice. Sigma1 Financial is focused on working with investment professionals, and content on this blog is directed at investment professionals, not individual investors. Please note that most or all of this content is raw, unfiltered, and ill-suited to individual investors. Further, no claims are made about the accuracy of data contained herein; all data is unaudited. Investment professionals should independently verify content on this blog before considering using in any way to make or influence investment decisions.