, , , , , , , , ,

As most of my readers know, I am currently stuck in compliance registration waiting to be allowed to manage clients’ money again. The paperwork should clear soon, so until it does, I’ve been focused on developing the tools I will need to help future clients achieve their investment goals.

A new client recently approached me with an interesting mandate I’ll explain, and hope readers will discuss in comments to help me improve the strategy.

He cashed out of the market near the top in January 2008 and has watched the last few years with his nest egg safe but sidelined. He realizes if his nest egg is going to remain one, he needs to put his capital back to work. Given current market status, he expects a pullback and finds himself torn between market risk and market exposure. I have spent the last month working on a solution for these investment requirements.

He asked me to build a model that would be less volatile than the indexes, with an absolute return strategy for his account. He could underperform a bull market but wants nothing to do with a 20% pullback.

The turtle approach to alpha. Said differently, how do you generate alpha without beta?  I went to work and developed a lower beta model generating index-beating results.  In doing research, I looked back to the work of Alfred Jones, the father of the modern hedge fund. He was a reporter who was studying the market, and decided to pair two styles together. When combined, his strategy generated market-neutral results.

The Jones Hedge Fund Strategy was in business for 32 years, with 29 of them profitable. The company is still around as a fund of funds using hedged equity as its strategy. In 1968 Institutional Investor published this piece on the first hedge fund shop.

“The logic of the idea was very clear. It was a hedge against the vagaries of the market.”

He used a hedged pair strategy. The idea is you go long a quality stock and short an equal dollar amount on a like stock that has worse fundamentals. A hedged pair of like stocks should be market-neutral.

That is, if the equity markets go up 100 S&P points, or down 100 S&P points, the only difference in account value is the change in value between the pairs. The pair can go up in value together, or down in value together, or bifurcate in their returns, you only care about the rate of change between the pairs. This is the basis I have expanded on with my new model.

The JHB Capital Long/Short Market-Neutral Strategy.

The strategy is built around the idea of using natural pairs – Long Coke/ Short Pepsi or Long UPS/Short FedEx. The idea is to capture the spread between two similar stocks. Historically, the failure of hedged pairs has been hedges that didn’t work when most needed.

The premise behind the model is to select highly liquid pairs of stocks that move with the market noise as one, while slowly diverging away from each other. This generates a portfolio with little correlation to the market indexes as a whole.

The model is currently built using Excel files.  Samir, of www.investexcel.net, has graciously allowed me to have access to his own work, to morph it into what I needed it to do. He has built a number of excellent, free, and easy to use Excel tools which I highly recommend to anyone.  I have used two already for this project and hope to move the model to a serious database to track the ratings changes over time.

The model currently uses a “Stock Comparison” spreadsheet available at the link. I modified the sheet to incorporate a two-stock evaluation, generating a unique score based on the comparison of the Fundamental & Technical Analysis data.

In simple terms, I am comparing two stocks, and on each line item the model generates a 1 or 0 value. The pairs are compared via 40 different evaluations, some of which are true fundamental evaluations and others which are technically focused on the strength of the stock in question.

This gives us a dual view of both types of evaluations and generates a quant-based factor score for each item. Once added up, this gives each pair a moving ratio score which will be used to generate exits from the portfolio as well as hitting gain targets or stop losses.

The score will be compared against a sub-sector baseline generated from the “Stock Screener” available at the link which will be used to generate a mean score for each of the stocks in each sub-sector. This gives us a baseline to compare our individual company scores with the paired trades, to generate the best pairs of like stocks.

The portfolio will use a built-in exit on any pairs that have lost 1.5% of the account’s AUM. The portfolio will also take profits at specific return levels in a pair, allowing new pairs to be plugged in to redeploy the capital in question. Finally, one of the strategy’s goals is to evaluate the pairs allowing for a change in them, if the internals have changed significantly since the pair was initiated.

While I’m still developing the model, it’s advancing nicely and has begun generating some interesting results. Here is a snapshot of the current model, as backtested to close of market on December 31, 2011.

This draft of the model has deployed only 49.3% of its capital, with the balance in cash.

These results are for a basket of 9 paired stocks with equal dollar value on entry. The model has not generated any profit or loss sales yet. I’m still working on incorporating the paired value scores, which I will document in a follow-up post in a few weeks when the next upgrade to the model is deployed and generating results.

I would like to thank both Matt Busigin (@mbusigin) and Ryan Prociuk (@ryanprociuk) who were both generous in helping me with the proof of concept pieces necessary to build a working version.

I will be following this post, with others that focus on either the design or the results of the model.



  • Stock Screener here
  • Stock Comparison link