Balhiser Financial develops portfolio-optimization software designed to revitalize how professional investors build, refine, update, analyze, and test financial portfolios. We build software to seamlessly support innovative and disruptive risk models to mitigate, manage, and reduce risk — however you chose to define and model it.

The mission of Balhiser Financial is to empower you, the financial professional, by making your job more fun by automating the tedious bits — making you more productive, and freeing you to work on other priorities. Balhiser Financial performs semi-variance optimization using advanced proprietary algorithms. The Portfolio-Optimization Suite generates an efficient frontier (EF), allowing you to trade-off risk versus expected return. Moreover, tools such as Smart Tuner compute the exact trades to move an existing portfolio close to a chosen point on the EF with the fewest trades possible — keeping churn to an absolute minimum.  And tools such as User Tuner make it easy to explore risk/reward “what if?” trade-offs, such as holding a recommended-sell position to avoid a short-term capital gain.

Like a CPA, Balhiser Financial works with Investment Advisers, but is not an Investment Adviser. Investment advisers provide Balhiser Financial with portfolio asset data, and our software crunches the numbers and returns the results in a model very similar to a SaaS model.  All investment decisions are the sole responsibility of the client (the investment adviser), not Balhiser Financial.

Disruptive Technology: Rethinking Risk Models

The core product invites the investment profession to expand their view of risk. Risk does not have to be coerced into a single number, standard deviation (sigma, σ).

Investment adviser reps seldom quantify the expected “risk” of a portfolio — say, 15.3 percent per annum.  They may say little about risk, other than this is a “moderate risk” portfolio suitable to the client’s indicated risk tolerance (obtained by a 10-question survey).

Our company is not necessarily trying to change how investment risks are explained to clients.  We are, however, definitely challenging the professional investment community to mitigate investment risks, for their clients, with more advanced optimization and analysis tools.  This holds for external clients, and internal clients (aka proprietary trading).

First Disruption Concept: Alternative Risk Models

Briefly back to the classic risk model, standard deviation, or its squared cousin variance.  These numbers standardize risk, and these standardized risk models don’t do well with so-call “statistical outliers”, such as the flash crash of 2010, or the AP Twitter flash crash of 2013.  How do you feel when your stop-loss orders get triggered and 20, 30, or 40 percent of your returns evaporate in seconds?  Do your standardized risk models soften the blow?  Even your one-day VaR probably did you no favors.

Maybe you would have considered stop-loss (limit) orders!  They may NOT have triggered, and that could be a good thing.

There are investment risks, both old and new.  From sources such as old-fashioned market panics, to old-school program trading, to cutting-edge high-frequency trading algols, to unforeseeable black swans.

The risk model that is steadily gaining traction within the investment community is semi-variance (SV, or σd).  There are several flavors of semi-variance, however in my experience they tend to all work similarly.  When it comes to simple portfolio construction using diversified ETFs, such as VTI, VXUS, AGG, the difference between V and SV optimizations tends to be subtle.  However, for more complex portfolios containing individual stocks (and/or sector ETFs and/or country-specific ETFs), the difference between V and SV becomes strikingly clear. Balhiser Financial has had the privilege of working with and improving several well-designed portfolios. Based on our observations to date, semi-variance reduction produces measurably superior results versus variance-reduction approaches.

In many cases, portfolios optimized based on semi-variance minimization produce superior ex post variance results, in addition to superior ex post semi-variance results (than the exact same set of assets and min/max constraints optimized using MVO.)

Consider this:  If you can model risk, you mitigate it.  If you can imagine it, you can model it. And, with the right risk-mitigation software, you can quantitatively model, manage, and mitigate risk.

[notice]Past performance does not guarantee future results. “Optimized” portfolios may lose value, and may lose more value than pre-optimized allocations.[/notice]

The Killer App:  Good Analysis Plus Portfolio-Optimization

There are some (rare) analysts, who provide significant positive ROI. Great stock analysts not only find “hidden gems”, they identify good buying and selling opportunities for stocks that are household names. However, most individual analysts specialize, for instance, in sectors and/or geographic regions. Investment companies that have the benefit of several good and/or great analysts need a way to turn a set of stock (bond, ETF, alternative investments, etc.) recommendations into a robust portfolio. The killer app for investing is solid analysis working in harmony with advanced portfolio risk-mitigation technologies.  

Return: Yeah We Optimized That Too

The Portfolio-Optimization Suite Engine optimizes for reduced risk and increased expected return concurrently.  Essentially, our Portfolio-Optimization Suite finds the asset allocation mix that, for a given return, reduces non-systematic risk to a minimum.  Then it repeats the process for other required expected returns, building the efficient frontier for a given set of assets.

However another was of looking at it is that the Portfolio-Optimization Suite finds the maximum expected return for a given level of risk… rinse and repeat for various levels of risk.

The concept is simple:  Optimize for return versus risk.  Then graph, and interpret. Then comes the art:  Pick the best portfolio for the client and make the appropriate trades to get there (or close to there with minimal trades).

Mixing Evolution, Revolution, and New Challenges

Initially,  from Q2 2012 through Q3 2013, roughly half of our partners and clients were primarily interested in classic MVO. However, since Q4 2013, this mix has shifted towards a 2:1 split in favor of non-MVO optimizations. So far (as of 5/9/14) all the non-MVO optimizations have been modified semi-variance optimizations (MVSO or just SVO).

In general, I’ve noted that companies with larger assets under management (AUM) lean strongly towards SVO, while smaller companies still tend to favor MVO.

The current mission of Balhiser Financial is to utilize both MVO-based and non-MVO-based revenue to finance alternative-risk mitigation technologies.  Today, we favor semi-variance portfolio optimization. Tomorrow, we may recommend using some form of max-drawdown risk modeling and mitigation. That is the single greatest advantage of the Portfolio-Optimization Suite Engine… if a time-series-based risk model can be coded, HALO can optimize it.

So far every risk-model and data set we’ve thrown at the Portfolio-Optimization Suite Engine has produced robustly-optimized results. Occasionally, a data set or enhancement requirement has initially highlighted either run time or robustness issues with the HALO Optimizer, but in all cases to date (as of 5/9/14), I’ve been able to enhance and improve HALO to deal with it…. usually in 2 days or less.

Balhiser Financial, LLC is always looking for the next challenge.  While we hope your unique requirements and specifications will work seamlessly with the current version of Portfolio-Optimization Suite, we also enjoy, Yes I said enjoy!, the occasions when a new challenge is discovered.  These events only serve to make Portfolio-Optimization Suite better.  (And, to any software engineers that might be reading… Yes, we keep a diverse and growing set of regression tests to ensure that improvements are not just isolated optimizations; they are general-purpose improvements that are, at a minimum, net neutral over the regression test set, and, in most cases, are net positive.)