Opening Cross: Oh, the Weather Outside Is Frightful...

… Though the snow is quite delightful,
At least for a day or so,
But now make it go, make it go, make it go...

Good news: that’s probably my last display of poetic genius until next Christmas and New Year. Bad news: more rhymes likely this time next year!

But bad news for some may be good news for others (twisted individuals they may be…). The question is, how do you determine the difference between good and bad news? Is it merely in the eye of the beholder—a subjective determination based on other factors, such as whether your overall portfolio is receptive to or at risk from a certain event—or can it be mathematically quantified in a way that allows trading algorithms to make sense of news and reliably act on it?

This is one of the issues addressed by a roundtable of professionals involved in generating and processing news data, which you can read in full later in this week’s issue. Indeed, one of the problems with using news in computerized trading strategies is quantifying how great an impact an event or a news item will have on the price of an instrument. For example, when a stock goes ex-dividend, it’s easy to predict how the price will change as a result. Or when a company misses or exceeds earnings estimates, historical correlations can guide future assumptions with some degree of accuracy. But other types of news—specifically, those that involve any notion of sentiment—can be more problematic.

For example, if I write that vendor A sucks, how does one quantify what impact that would have on vendor A’s share price? Does it depend on factual events that denote “suckiness”, such as poor margins, implementation backlogs, or poor customer service ratings? Or on the tone of language used—such as if I call the company “disappointing” instead of “sucking”? (And would an algorithm even know the meaning of “suck” in this context?)

Would it have the same influence if the comment is in an article predominantly about another vendor, and mentions Vendor A only in passing—and would that comparison have a correlated, positive impact on the other vendor by association? And how much of a difference does it make if I say it in a news story versus an opinion column, or if I’m stating my own opinion or quoting a third party versus quoting an executive from the company in question (like former UK jewelry chain boss Gerald Ratner calling his products “crap”)?

Sentiment is where perhaps the greatest difficulty lies, but which perhaps also offers the most opportunity. The areas addressed so far by machine-readable news include economic releases, corporate financials and other unstructured data that can still be reduced to numerical values or yes-no equations. While I don’t want to call this easy, it is still the low-hanging fruit, and the markets are barely scratching the surface of potential uses.

Applying the speed of computers to event-based trading could yield huge gains, but failing to appreciate the finer points of a factual event—or the way in which it is reported (“Despite strategic push, Vendor A still fails to make profit” versus “Vendor A’s brilliant strategy means losses are less than expected”) could be the difference between a winning and disastrous strategy.

Having spent all these column inches pontificating about news, I can’t help but realize that all four items in this week’s Scrolling News section involve indexes and exchange-traded funds. Perhaps that’s just a coincidence.

Or perhaps it reflects that no single content set or data type owns the future of investment and trading, and that successful strategies will have to incorporate not just prices or news, but other inputs such as indexes as well.

Here’s to 2011 and the innovation that this New Year will bring!