August 15, 2014

The Alpha Factor: Helping Traders Discover Hidden Value

by Stephen Malinak

Though faced with increasingly complex databases of information, today’s buy-side and sell-side equity traders still can’t afford to miss any hidden indicators. Traders who consistently succeed at this are justly celebrated for their ability to “separate the signal from the noise”. But with new signals and increased levels of noise hitting the street every day, differentiating between them becomes a bigger challenge with every market open. Portfolio managers expect traders to add more value than ever, from smart execution, to identifying new sources of alpha, to suggesting trade ideas, to helping build custom indices. Add the ever-shifting regulatory environment, and traders confirm that access to the best tools, models and information is increasingly important in dividing those who seek alpha from those who actually find it.

More data demands more monitoring, across every link in the investment chain. Company analysis (fundamental, technical, and a host of niche quant measures), corporate access, insider transactions, government filings, credit analysis, earnings forecasts, manager strategies, institutional holdings, analyst recommendations, investor sentiment, regulatory changes, macro economics, political shifts, and so much more, can either offer real signals or more noise. The best traders keep a sharp eye out for anomalous information that either validates a new opportunity or invalidates an old one, even as they work on trades spanning a dizzying range of execution choices.

But there’s good news. To take an “uncorrelated” analogy from science, chemicals that are inert on their own often form volatile compounds when combined with other chemicals. Similarly, investment data might seem unimportant in isolation, but when intelligently combined with another data set it can catalyze highly profitable trading ideas.

Thomson Reuters StarMine models provide good examples of this. Important trading tools for predicting earnings surprises, they also offer uniquely sophisticated screening factors for traders looking for innovative ideas, and reveal hidden value that might otherwise stay hidden.

Our approach originated with a unique set of data we’ve been collecting and analyzing for over ten years: the earnings estimates of leading analysts, segmented by degree of accuracy. We discovered that analysts who have consistently been the best at predicting earnings are nearly four times as likely to remain the best in future, while analysts who have been least accurate are roughly four times as likely to maintain their inaccuracy. By strongly weighting the best and accounting for estimate “age”, we generate “SmartEstimates®.” We then measure the difference between those numbers and the consensus, highlighting differences greater than 2% as “Predicted Surprises.” These turn out to be directionally correct 70% of the time – in other words, if a Predicted Surprise is greater than 2% to the upside, 70% of the time company earnings will come in above consensus (the reverse is also true).

Portfolio managers hate surprises, so in preparation for an expected earnings call, traders turn over every rock in search of potential shocks. Using our models to test a single company or a basket of companies, they can turn over many of those rocks at once — including more than a few that other traders may not even know about.

Over ten years ago, we started by focusing on analyst revision models because they were new and powerful. These days we track so many other data sets that by combining our best models we keep developing new ones that are greater than the sum of their parts. As an example, we have very effective value models that separate cheap stocks from overpriced stocks, and other models built on the inertial tendency of trends to continue into the future. By combining them, we developed a third – StarMine Val-Mo — that identifies cheap stocks poised for rebounds, and over-priced stocks poised for reversions, and is finely-tuned enough to differentiate between “value traps” and stocks that are truly undervalued. That’s exactly the sort of insight traders look for.

Even at the most pragmatic level, factor-based models are directly useful in today’s trading environment. As an example, we’ll close with this hypothetical scenario:

Europe is underperforming. Your portfolio manager has been careful to focus exclusively on U.S. healthcare, and believes he’s successfully isolated his holdings from the crisis in Europe. As a trader though, your job is to take nothing for granted, and distrust anything you haven’t personally looked into. Using our StarMine models you check for possible signs that the European situation might still pose a problem for your PM. Sure enough, one model reveals that a large position has credit exposure with an Italian bank. Another indicates recent selling activity among company executives. Yet another uncovers a large holding with a big negative Predicted Surprise and a serious inventory buildup. This type of screening could potentially save/make that PM millions of dollars.

Clearly, factor-based models such as ours can be as useful to traders as they are to quants and portfolio managers. For more information about the StarMine models and Thomson Reuters Eikon, visit

Receive stories like this to your inbox as they are published. Subscribe here and follow us @Alpha_Now on Twitter. If you are looking to access Thomson Reuters data or analytics, register for a free trial.

Article Topics