Are prediction markets accurate?

"Prediction markets are more accurate than the polls" is a headline you have probably seen. It is a fair claim in some cases and misleading in others. The honest answer is that prediction markets are usually well-calibrated, sometimes beat other forecasting methods, and occasionally get things badly wrong. Understanding when they are accurate matters more than treating them as a crystal ball.

If you are new to the format, our primer on what a prediction market is explains the basics. This post is about how much you should trust the number.

What "accurate" even means for a market

A prediction market does not promise to tell you what will happen. It gives you a probability. A market trading at 70¢ for "Yes" is saying the event is about 70% likely. So how do you grade a probability?

You cannot judge a single 70% forecast by whether the event happened — a 70% chance is supposed to fail 30% of the time. Instead you look at calibration across many markets. Take every market that closed near 70%. If prediction markets are accurate, roughly 70% of those events should have actually occurred. When forecasters check this, well-run markets tend to be impressively well-calibrated: things priced at 30¢ happen about a third of the time, things priced at 90¢ happen about nine times in ten.

That is a higher bar than most pundits clear. A commentator who says "this will definitely happen" is right or wrong; a market that says 82¢ can be right about being uncertain.

Why the price behaves like a probability

The mechanism is simple. If a market looks too cheap — say "Yes" is trading at 40¢ but you have good reason to think the real chance is 60% — you can buy and expect to profit. Enough people doing that pushes the price up until the easy edge disappears. The price settles where the money, on balance, stops disagreeing.

This is the wisdom of crowds with a twist: people are not just voting, they are risking their own money. A confident, well-informed trader moves the price more than a casual guess, because they are willing to stake more on it. That weighting toward informed conviction is a big part of why the numbers hold up. If you want to practise turning a price into a probability, our guide to reading the odds walks through it.

Where prediction markets go wrong

Accuracy is a tendency, not a guarantee. Markets fail in predictable ways:

• Thin markets. A question with almost no traders and tiny volume is just a few people's opinions wearing a probability costume. Low liquidity means the price can be stale or easily nudged.
• Longshot bias. Very unlikely outcomes often trade a little too high, and near-certainties a little too low — people overpay for the thrill of a big payout, the same pattern seen at racetracks.
• Correlated blind spots. If everyone reads the same sources, the crowd can be confidently wrong together. Markets aggregate information; they do not manufacture it. Garbage in, garbage out.
• Ambiguous resolution. If the rule for settling "Yes" or "No" is fuzzy, the price reflects the argument about the rules as much as the event itself.

None of these break the idea. They just tell you which numbers to trust: deep, liquid, clearly-defined markets on well-covered events are the ones that earn their reputation.

How to read an accuracy claim sensibly

When someone says a market "called it," ask three questions. Was the market liquid, or a curiosity with a handful of trades? Did it say something bold, or was it hovering near 50/50 where being "right" is a coin flip? And are we counting the misses too, or just remembering the dramatic hits? A forecasting method has to be judged over a long run of predictions, not one viral screenshot.

Done that way, prediction markets hold up well. They are not always right, and they were never supposed to be. What they offer is a single, honest, continuously-updated number that already digests the news, the polls, and the arguments — and that moves the moment the facts do.

The best way to build intuition for this is to watch prices move and see how often the confident ones pay off. On Outcomer you can do exactly that with virtual money — no risk, just practice reading the crowd and testing whether you can spot when the price is wrong. Give it a try and see how well-calibrated your own hunches turn out to be.