Clarity Strategic^TM

Recently, my good friend Don Ake gave a presentation to the local chapter of the Institute of Management Accountants about "unusual economic indicators." In this talk, he introduced his audience to such useful tools as the Short Skirt Indicator, shown here. :-) And he neatly summarizes all the reasons that forecasting the value of a stock, or of a stock index, is difficult. I decided to collect all these reasons and present them here, because it's not just stock we predict.

• weather
• sports team (and individual) performance
• start-up (and mature) company performance

So why is forecasting so hard?

(1) Short-term fluctuations. If you look at the performance of a stock index over time, that tends to be noisy data. Those fluctuations can look significant, if you only look at the near-term.

If we react to those fluctuations, it means we have been caught up in Tuffy Rhodes Syndrome. Tuffy Rhodes is a baseball player, an American who has had a long career in Japanese ball, where he has played with distinction.

But what many Americans remember about Rhodes is for his performance for the Chicago Cubs on Opening Day, 1994. On that day, he hit three home runs off Dwight Gooden in his first three at-bats. Fantasy baseball players notice performances like that because they are... well, fantastic. I traded for Rhodes on opening day, figuring he had a leg up on hitting 30 homers for the season. He hit EIGHT for the season and was released in 1995. What did I give up to get Rhodes? It doesn't matter. it was a bad deal, and I made the deal based on the noisy near-term data, rather than looking at his whole career.

We make the same mistake in looking at global warming: either by assuming that there's no such thing because it's snowing outside, or by assuming the world will end tomorrow because today it's 95 degrees. If you think about such mistakes, it seems like only foolish people would make them, but otherwise intelligent people make mistakes like these every day.

If you want to predict the future, you have to take a longer view. You need a lot more time, and a LOT more data.

(2) The effects of the unexpected trend. One thing that happens with out predictions over a slightly longer time period than the short-term noise is a trend based on something that doesn't happen every day.

One example of this is (turning to baseball again) the Indians' 30-15 record to start the 2011 season. Another would be the rise of the flash mob, which most of the world did not predict would lead to a sharp increase in viral videos.

If we react to early to the unexpected trend, we can't take advantage of / protect ourselves from its effects in the longer term. In either case, we have been victimized by All Hell Breaks Loose Syndrome.

You remember how, during "Harry Potter and the Deathly Hallows, Part 2," Harry told Hermione this: "we plan, we get there, and all hell breaks loose." And this is what you must assume will happen. Expect it, even if you cannot guess the effects it will have on your forecasting.

This is a comparison of Ake's Model-T stock trend forecasting model against the performance of the Standard & Poor's 500. Don's model fares well here, though it picks up on the trends a little earlier and the bad times tend to be much worse. There's no noise visible here because this is about 20 YEARS worth of data. With stock value, as with baseball performance, there's plenty of data there to check your guesses against.

(3) Outliers. These are those occasional points that end up far away from the rest of your data. The temptation is to react to them. But to do so would be a mistake; we can't guess what they'll be responsible for. Nor can we know what causes them unless we look very closely.

There are two possible mistakes here: to assume there is no outlier when there may really be one, and to assume there is an outlier where there really is not.

One example of this (turning back to baseball yet again) is the popular perception that Yankees shortstop Derek Jeter is a "clutch hitter." Because he has had some big hits with the game on the line, people believe that somehow Jeter is a different player in those situations than at any other time. Yet his statistical performance says that's just not true.

And so we have to be careful, in case what seems like an opportunity is coming along. If we believe that Jeter - or whomever else - is a clutch hitter, for whatever reason, then we might well assume that The Muslim we see boarding a plane is a terrorist, though we have seen no evidence. This is us ignoring Clutch Hitter Syndrome.

Summary

• Tuffy Rhodes Syndrome - decisions made with yesterday's data and no consideration of history
• All Hell Breaks Loose Syndrome - decisions made without considering what could go wrong, or even what may be going wrong right now
• Clutch Hitter Syndrome - using a single case to assume general behavior

Why do people make mistakes here, anyway?

• Because sometimes we have an (unproven) idea and we won't let go of it.
• Because sometimes the homework we must do to improve our decisions is just too hard.
• Because sometimes we just plain want the answer, right NOW.

Be careful out there.