Thursday, 1 March 2018

The Rule of 40%

There was an embarrassing error in the Wall Street Journal’s article on pundits and the strategy of the 40% probability claim. 

The full article is below, but first see if you can spot the error in this section:

Citigroup’s Mr. Suva forgot to hedge. He put a 40% chance on Apple buying Netflix, but also a 25% chance Apple buys Walt Disney Co. , a 10% chance each it buys one of three videogame makers, and a 5% chance it buys Tesla Inc. That sums to 100%, implying it is mathematically certain Apple buys one of them.
How Do Pundits Never Get It Wrong? Call a 40% Chance

Talking heads have learned that forecast covers all outcomes; ‘I just said it was a strong possibility.’

What are the chances that readers will make it to the end of this article? About 40%.

If you do make it, that prediction will look smart. If you don’t, well, we said the odds were against it.

Such is the nature of the 40% rule, a favorite forecasting tactic of Wall Street analysts and other prognosticators trying to make a bold call without being too bold.

Former British Prime Minister Tony Blair said last month there’s a 40% chance that Brexit will be reversed; Citigroup Inc. analyst Jim Suva wrote that there’s a 40% chance Apple Inc. buys Netflix Inc.; and Nomura Holdings Inc. economist Lewis Alexander said there’s a 40% chance Nafta gets ripped up.

The nice thing about 40% is that you never have to say you were wrong, says Peter Tchir, a market strategist at Academy Securities. Say you predict the Dow Jones Industrial Average has a 40% chance of hitting 30000 before year-end.

“Get it right and you can say ‘See, I was telling everyone it could happen,’ ” he says. “Get it wrong and you can weasel your way out: ‘I didn’t say it was likely, I just said it was a strong possibility.’ ”

With a 24-hour news cycle, outlets from cable channels to newspapers are always looking for an expert to weigh in. If they offer an audacious estimate that will get clicks, all the better. The trend has boosted the industry of analysts and talking heads who predict everything from election outcomes to corporate earnings.

A whiff can ding a forecaster’s reputation.

Nate Silver became America’s most famous election forecaster when he called all 50 states correctly in the 2012 presidential election. Four years later he was criticized for repeatedly projecting that Donald Trump stood no chance in the Republican primary and for his final pre-vote projection that gave Hillary Clinton a 71% chance of victory.

Mr. Silver says he deserves some criticism for his primary projections, but not his general election forecast, which he called “highly informative and useful” since others gave Mr. Trump a smaller chance.

To protect their reputations, pundits hedge. They may not provide a date by which a forecast will occur. They often “cluster” forecasts together with other analysts around a “consensus” figure so that everyone will probably be the same amount of wrong.

Citigroup’s Mr. Suva forgot to hedge. He put a 40% chance on Apple buying Netflix, but also a 25% chance Apple buys Walt Disney Co. , a 10% chance each it buys one of three videogame makers, and a 5% chance it buys Tesla Inc. That sums to 100%, implying it is mathematically certain Apple buys one of them.

A Citigroup spokeswoman said Mr. Suva doesn’t believe a deal for one of those companies is guaranteed. She said the forecast was really a conditional probability, contingent on Apple using its huge cash pile for what she called a “mega deal,” which she said Mr. Suva actually views as less likely than a large stock buyback. The conditional probability wasn’t specified in Mr. Suva’s research note.

“Pundits and gurus master the art of going out on a limb without going out on limb,” says Philip Tetlock, a professor at the University of Pennsylvania who has made a career analyzing which people forecast well, and why. One of his pet peeves is how gurus use vague terms like “distinct possibility” instead of percentage odds when they describe probabilities. That makes it easy to wiggle out of, or take credit for a forecast, since it isn’t clear at all what a distinct possibility is.

But one drawback of percentage odds, Mr. Tetlock says, is that people are often unclear on what they actually mean.

Mr. Silver can relate. Some of his harshest critics took his 71% projection of a Clinton victory as a sure thing that she would win, he says. “You wouldn’t cross the street if there was a 30% chance you’d get hit.”

Courageous contrarian calls are the best way forecasters capture the public’s attention, and get television time. New York University Professor Nouriel Roubini was dubbed “ Dr. Doom ” for correctly predicting the financial crisis. Then in 2010 he projected a 40% chance of a “double-dip recession” in the U.S. It didn’t happen.

Mr. Roubini says he doesn’t remember the projection, but that he takes pride in sticking his neck out, as with his latest call that Bitcoin is the biggest bubble in history and will go to zero.

“I would not rule out that I’ve committed the sin of the 40% rule,” said Prof. Roubini. “Everybody has done so.”

“There’s an aspect of infotainment” that Wall Street forecasters always keep in mind, says John Kilduff, portfolio manager at commodities hedge fund Again Capital. In September 2015 Mr. Kilduff told CNBC viewers that crude oil had a 40% shot at falling to $20 per barrel. Then at $45, oil followed its downward trend before bottoming at $30.

“You’re always riding the hero-shithead roller coaster,” says Mr. Kilduff, “we all have plenty of haters, and they’re even more visible now with Twitter . ” He noted one correspondent who took a forecast personally. “Hope you lose EVERYTHING [on] your short,” the person wrote in a salty email reviewed by the Journal. The correspondent also called Mr. Kilduff ugly.

“The old 40% trick!” recalled Stephen Roach, formerly Morgan Stanley’s chief economist. “A warning of a looming forecast change, a rising risk assessment, a way to cover your rear—or a combination of all three.” Mr. Roach said that by stamping a 40% probability on a possible outlier he could call clients’ attention to shifting winds without changing his underlying forecast.

A nonrigorous examination of Mr. Roach’s past forecasts showed he put a 40% chance on two recession predictions in 2002 and 2004, and then another recession projection in 2010. All of those he got wrong, or rather, got right, since he said the odds were against.

The 40% rule can be useful for all manner of punditry. British boxer Anthony Joshua speculated in 2015 that underdog Tyson Fury had a 40% shot to beat heavily favored champ Wladimir Klitschko. Mr. Fury won, and now wants to fight Mr. Joshua. A spokesman for Mr. Joshua didn’t respond when asked what chance the boxer would give himself against Mr. Fury.

By the way, if you made it this far, we always predicted you might.

There was also another nonsense click-bait headline and article from the Yahoo!Finance pages this week.
The MarketWatch headline read:

Steve Wozniak had $70,000 in bitcoin stolen after falling for a simple, yet perfect, scam

Well, not exactly. At the time the Apple co-founder tried to sell his 7 Bitcoins, they were worth about $700 each, so the writer is being somewhat disingenuous in claiming that the loss was $70,000 when it was actually less than $5,000.

Then there’s a Mashable article on the same topic with the false headline:

Even Steve Wozniak said he fell victim to a bitcoin scam
No, he didn’t fall victim to a bitcoin scam. He fell victim to an old fashioned credit card scam, and that the object stolen happened to be bitcoins is irrelevant.
As for the probability error in this post's lede, anyone who has a basic understanding about probability will know that the probabilities of independent events do not sum as the authors suggest.

Using their numbers, I make it a 31.16% chance that Apple buys none of the companies listed, not 0% as the writer suggests.

I’m not suggesting that any bet listed in the example here would be the best value ever, but should Crystal Palace be quoted at 2/1 next season to win the Premier League, 2/1 for the FA Cup and 2/1 the League Cup, the probability that they will win one of the competitions is not 100%.

Someone miserably failed Basic Probability 101.

No comments: