My Worst Predictions

I’ve made my share of bad predictions. I’ve had lots of good predictions too. And being right often doesn’t matter, because in the end almost nobody will care if you were right or wrong, and if you’re going to go about placing bets based on your predictions, it’s incredibly difficult to get the timing right.

Some of my best predictions include my general bullishness on software and the internet since the 90s, betting on Airbnb, and also Bitcoin. In the case of the last one, I was right about my bet, but got the timing wrong.

In terms of my worst predictions, I have a few memorable ones. Most recently, I bet the Federal Reserve wouldn’t follow through with rate hikes (which, as you likely already know, it has). In fact, I predicted they wouldn’t get above 0.50% for the federal funds rate by the end of 2022. Woops.

Another one of my bad predictions (so far, I guess it could change) was that the internet advertising bubble would burst. We already know advertising is remarkably ineffective, and Freakonomics did a two-part series (one, two) on this. I still think advertising is a giant bullshit bubble, but it may never burst in the sense that big ad companies would go belly up or have huge revenue declines (Google, Facebook, etc). The thing about ads is that advertising are always competing for eyeballs, so if one company stops running ads, there are a bunch right behind them to take their place, so they can’t stop the ad spend because then their competitors would get in front of the eyeballs (or bots).

Another one of my bad predictions is that when Obama won the election there would be big changes, and he’d live up to his promises. In spite of the Democratic party having a short-lived supermajority, they largely fell short on much of what was promised. To give credit where it’s due, they did eventually delivery on a pared down healthcare bill with a few key provisions (notably banning insurance companies from denying coverage due to “preexisting conditions”, which is a dystopian political term for people who aren’t in perfect health).

Another prediction I made was that the COVID-19 thing was going to be another short lived news cycle, which turned out to be terribly wrong. As of today the NYT still shows their “Coronavirus Pandemic” widget on the front page, although they’ve moved it from the top to 2/3rd of the way down the page.

They’ve slowly nudged it down as interest in the subject has waned, but I guess it still gets enough clicks that they make room for it on the front page.

And certainly COVID is more than just a news cycle, but the NYT doesn’t bother to make room for the other health crises in the US: obesity, opioids, cars, pollution, etc.

Being right is only 10% #

I guess that brings me to my point: whether you have the right prediction or not is kind of irrelevant. What matters is how you act on it, and in the case of making bets (such as with stocks) it’s very hard to get the timing right. In fact, I’d argue that getting the size and timing right is 90% of the problem when it comes to markets.

Constructing a portfolio to bet on any one particular thing can be quite tricky, and it’s often not as simple as taking a short or long position. For example, if I wanted to bet on the demise of Facebook (which, I think a lot of people are also betting on) it seems like the easy thing would be to short META, or buy some puts, right? Wrong.

Herein lies the problem with betting on such things:

• Most of the time, such bets are too obvious.
• Markets are efficient and clever enough to know that lots of people think Facebook is a failure, even though it’s still quite big, generates a lot of revenue, and probably won’t disappear anytime soon.
• Most of the tools available to ordinary folks like you and me have already priced in the likelihood of Facebook’s demise, meaning it’s “priced in”, and even if you’re right you probably won’t be more right than the market makers.
• It can take much longer for things to play out than you can remain solvent for. Or, as they saying goes, “markets can stay irrational much longer than you can stay solvent”.

Some of the most clever fund managers or investors have managed to be successful with big bets like this in part by using specialized derivatives, or by constructing clever portfolios that protect them from the risk of decay or (in the case of shorting) interest payments. For example, if I want to short a stock like GME, I have to pay 9.6% (at the time of writing) to borrow the stock to short. You can do some fairly simple calculus to see how much the stock would need to fall and at what rate over some period of time to see what your return would be. When the volatility spikes, the cost to borrow (in the case of shorting) goes up, and so does the implied volatility on options (which makes the cost of shorting with options higher, too).

There are a lot of technical issues that go into this, but to get back to my point: the best fund managers figure out alternative ways to make such bets, which often involve some kind of spread (i.e., buying one thing and selling another) which can balance out some of the cost of shorting, depending on what you’re doing. Of course you can still be wrong and lose money, but the point is to limit the max loss.

Minimax #

And now we arrive at my last thought which is the idea of applying minimax to such bets. Minimax is short for minimizing the maximum loss (or the reverse, depending on which way you’re betting), and it’s basically the algorithmic equivalent of the Nash equilibrium, popularized by that film with the angry Australian Russell Crowe.

The Nash equilibrium basically says that if all players in a game act in everyone’s best interest, then you’re most likely to have a good outcome for everyone. It’s kind of related to prisoner’s dilemma, in that if both prisoners cooperate, they will have a decent outcome.

Anyway, getting back to the subject of making predictions and betting on them: if you do decide you want to construct a bet, it’s important to try and construct a bet in such a way that you always minimize the max loss. I think people might have a tendency to try and take bets with the highest possible returns, which I supposed would be maximax, but from a strictly statistical or game theory perspective, that is a losing strategy. Or rather, you might occasional be right, but most of the time you’ll be wrong, and in the end you’ll probably lose.

There’s also another way to think about this, which is that instead of placing bets to profit from your predictions, you adjust your bets to decrease risk. Probably the simplest (or most common) example of this is the application of Modern Portfolio Theory to investing: you can minimize your downside risk by diversifying across non-correlated assets. Of course, this doesn’t work so well when everything becomes correlated, as we’ve seen recently in a rising rates environment.