Automated Driving on Polymarket

A tour of missed opportunities

This piece is the result of an experiment: Substack recently added the ability to embed live Polymarket markets in posts, and I wanted to see what I could learn from them. The results became an issue in its own right.

---

Bets are a tax on bullshit.

I use that term deliberately, even though it’s coarse, because it has a specific meaning, one that nonsense or lies or malarkey doesn’t convey. As per the philosopher Harry Frankfurt, bullshit has a precise definition: it is speech produced without regard to its truth or falsity. It doesn’t intend to track reality, but to persuade, impress, or signal some quality in the speaker. Whereas liars aim to conceal the truth, bullshitters are indifferent to it. They are playing a different game.

Bullshit is everywhere, and it makes it hard to understand the world. This is a problem, and a bet is one solution to it, because a bet clearly aligns speech with belief. Someone might say that he’s sure Canada will win a gold medal in shot put at the next Olympics. Does this person really believe that? He might be bullshitting, saying words that don’t correspond to anything about what will happen at the Olympics, but instead correspond to how much he loves the country… or wants us to think he loves the country. Ask him to wager $100 on whether Canada wins a gold medal at shot put, and his incentives change: now, his claim must rest on what he actually thinks will happen, not what he wants us to think about him. If he declines the bet, you have learned something about what he really thinks.

Prediction markets are an attempt to scale up this insight.

Polymarket hosts well over a hundred open markets on innovative mobility, and their prices are telling. In the Tesla-adjacent markets—which dominate the platform, in our domain at least—a large portion of the crowd is systematically pricing technology risk when it should be pricing regulatory risk, and the result is a consistent overestimate of near-term Tesla delivery.

How Prediction Markets Work

As per Scott Alexander, who made the case at length in his Prediction Market FAQ, a price on a prediction market represents the collective judgment of everyone willing to put money behind their beliefs. If the market underestimates the probability of something, anyone with better information can profit by betting Yes, and in doing so, move the price toward accuracy. Any persistent mispricing is a standing invitation for someone who knows better to make money.

This self-correcting mechanism is what makes prediction markets, at their best, epistemically useful: they reward people for being right, and penalize them for using speech as performance.¹ In that spirit, here is what I think the markets are getting wrong, and why.

Before I tell you that, though, let me say a few things as clearly and strongly as I can:

● I cannot trade on Polymarket; Ontario law prohibits me from doing so

● Even if it didn’t, I wouldn’t; I don’t hold mobility-related equities, because I don’t want my analytical work to be influenced by my portfolio

● There is no financial upside for me in the following analysis

● NONE OF THIS IS ADVICE TO YOU ON HOW TO GAMBLE, and if you do, I disclaim all responsibility

You may think that my own lack of skin in the game gives you good reason to discount everything I’m about to say. The only response I can make is that I do care about the truth for its own sake, and that perspective informs all my work at Changing Lanes.

Another thing: the markets never stop moving, but text is frozen in time. My analysis reflects the analysis at time of drafting this piece, which accounts for any difference in the price I discuss and the market you see.

Five Open Markets

Will Tesla launch driverless robotaxis in California by June 30? (17% chance)