When Prediction Markets Got It Wrong

Prediction markets have a great track record — on average. But averages hide nights when the favourite lost and everyone asked: "so much for the wisdom of crowds". Here are the famous cases, and an honest verdict on each.

First, what counts as "wrong"?

If a market gives an outcome 25%, it should happen about one time in four. When it does, the market wasn't necessarily wrong — it said this could happen. A market is genuinely wrong when it's systematically miscalibrated (its 25% events happen 50% of the time) or when it ignored information that was available. Keep that in mind as we go. Our guide to implied probability and calibration explains the idea in depth.

Brexit, June 2016

On the night of the UK's EU referendum, betting markets had Remain as a strong favourite — often around 75–85% as polls closed. Leave won 52–48.

Verdict: a real miss, partly. Polls were close throughout; markets were more confident than the polls justified. Analysts later pointed to large Remain bets from London-based traders and a tendency to assume the status quo wins. The lesson: who is trading shapes the price.

Trump vs Clinton, November 2016

Markets gave Donald Trump roughly 20–25% going into election night. He won.

Verdict: unlikely, not impossible. One-in-four outcomes happen. Markets were in line with most forecasters, and better than some models that gave Clinton 90%+. Still, the miss changed how traders think about polling errors that move in the same direction across many states.

The 2020 "red mirage"

On election night 2020, Trump briefly became the favourite on some markets as early in-person votes were counted before mail ballots.

Verdict: overreaction. The pattern — in-person votes first, mail votes later — was widely predicted. Traders who understood how votes were counted bought Biden cheaply. The lesson: know the process, not just the headlines.

2022 midterms: the "red wave" that wasn't

Markets expected Republicans to take the Senate. Democrats held it.

Verdict: a shared miss. Pundits, some polls and markets all leaned towards a bigger Republican night. Markets didn't add much beyond the consensus.

Sports: the inevitable upsets

Every season, 90% favourites lose. That's not failure; that's what 90% means. Over thousands of games, sports markets tend to be very well calibrated — though, as a Bloomberg analysis noted in October 2026, traders still pour money into longshots that lose far more often than their buyers seem to expect.^[2]

What these misses teach

1. A probability isn't a prediction. 75% means "likely", not "certain".
2. Herds exist. When most traders share the same assumptions, markets can be overconfident.
3. Process beats vibes. Knowing how results are counted or resolved is an edge.
4. Liquidity matters. Thin markets are easier to push around — see prediction market liquidity.

Despite the misses, decades of research suggest markets usually beat or match the best alternatives.^[1] For why that works, read wisdom of crowds.