Predictions and Black Swans
What is a prediction? How does one evaluate whether the prediction was accurate? What is "edge" and how does it differ from a specific forecast?
These are questions that every market participant should understand. Few actually do.
Extremely Unlikely Events
Eddy Elfenbein at Crossing Wall Street is one of my colleagues at TheStreet.com’s RealMoney site. Eddy provides interesting perspectives on the market and many research posts highlighting observations not noted elsewhere. Eddy (now added to our list of featured sites) has been on our "daily read" list for a long time. This week he raises an interesting question about predictions and the market, using one of our favorite comparisons — sports. While we maintain that investing is not gambling, there are many reasons to use the analysis of gambling to study risk/reward in investing. Most significantly, there are more predictive opportunities, so the study of results is more robust.
Eddy considers one of this week’s NFL matchups, the hapless Dolphins playing the apparently invincible Patriots. He notes the similarities between investment markets and the free market in wagering on NFL games, observing that there are not many similar cases.
This is a good "Black Swan" lesson. The Tradesports.com futures contract for the game, a price freely negotiated in a free market, makes the odds of a Dolphin victory to be about 13-1. Let us suppose that your own analysis showed that the "correct odds" were only 9-1. Making an investment in Dolphin futures has two characteristics:
- It has significant edge.
- You will probably lose on this specific occasion.
Black Swan bets are long-range propositions. One cannot expect to win on a specific occasion, but the advantage over time is huge — assuming the analysis is correct. The point is that we can never evaluate the system based upon a single choice. It can only be judged after many similar situations have occurred.
Finding Black Swans
One illustration of the approach to extremely unlikely events is an annual exercise by Doug Kass. Each year he comes up with an out-of-the-box list of predictions that are non-predictions. They are non-predictions in the sense that any particular example might seem unlikely. By listing it, Doug suggests that it might happen, with odds greater than the market might believe. If the investor can find a way of playing for these events, it might be a cheap shot. His record this year has been very good.
The market also has problems in dealing with events that are unlikely, but have a more substantial probability than the black swan. We challenged our readers with an example designed to make them think. Foremost among these is the chance of a recession, and the accompanying prediction of a significant market decline.
Please note that this is a more specific prediction. It should be accompanied by a time frame. Most economists use a one-year forecast. For the market prediction to work, the prediction must also have a market consequence. If the recession forecast is reflected in earnings projections and stock prices, one could be right on the recession question and wrong on the market.
There is always a chance of a recession. In a given year, the chances are about 20%. When economic growth is already at a reduced pace, the odds are higher.
Evaluating the Prediction of Unlikely Events
In past articles we have tried to explain the difference between a specific prediction and discovering "edge." Non-economist pundits frequently criticize the economic community for a failure to predict recessions. This is a serious error. Recession forecasting is probabilistic, not a 100% call.
A Forecasting Exercise
Economists (on average) place the odds of a recession in the next year at 35% or so. Some range higher and some much lower. Non-economists and the public put the odds much higher. Let us consider this using Eddy’s example of the NFL. There are several games where the odds of winning are about 20%. Let us suppose that you believed these odds to be 35%. You would have edge if making many predictions, but you would expect to lose each specific bet.
Here are the games where the odds are about 4-1:
Arizona versus Washington
Atlanta versus New Orleans
St Louis versus Seattle
San Francisco versus New York Giants
The question is whether any respectable pundit would predict a victory for any of these underdogs. He would usually lose and seem foolish, even if he had significant "edge" on the odds. One of these teams will probably win. (St. Louis?) Despite the likelihood that one will win, it would be poor form for a pro to pick any of these big underdogs.
A specific forecast has an outcome and a time frame. Even after one knows the outcome, it is impossible to know whether the forecast was good. It is one specific result out of many possible outcomes.
Investors who listen to those asserting certainty about a recession are making a mistake. No one can do this. Similarly, criticism of economists for using probabilities is wrong. Understanding the concept of "edge" and what is already reflected in market prices is crucial. Any trader or investor who does not understand this is operating at a big disadvantage.
Black Swan events are a bit different. These non-specific forecasts are relevant, yet not specific. The investor needs to consider ways to allow for these possibilities while not allowing them to determine the overall investment posture.