9 posts
Prediction markets are accurate where they are deep and short dated, and unreliable where they are not. On a cross platform sample of resolved markets measured to 14 October 2025, Polymarket scored a Brier loss of 0.1652 and Kalshi 0.1982, against 0.25 for always guessing 50 percent. On the 2024 US election specifically, one study found that only 67 percent of Polymarket markets priced the eventual winner above 50 percent.
6 min read
Beating a prediction market means being right where the price was wrong, often enough and on enough events that luck stops being the explanation. There is no trick that works everywhere. What works is a method: start from the price rather than your opinion, disagree only in the few situations where prices are known to be soft, put a specific number on it, and score every call against the price at the moment you made it.
10 min read
A Brier score measures how far your probabilities were from reality. For each forecast, take the probability you gave, subtract the outcome written as 1 for happened and 0 for did not, and square the result. Average that over all your forecasts. Zero is perfect, 0.25 is what you get by saying 50 percent every time, and 1 is as wrong as it is possible to be.
8 min read
Calibration is the match between what you claim and what happens. If you are calibrated, the events you call 70 percent happen about 70 percent of the time, the ones you call 90 percent happen about 90 percent of the time, and so on down the scale. It is a property of a batch of forecasts, never of one. A single 70 percent call that comes true proves nothing.
7 min read
A track record assembled from social media posts measures what survived, not what was said. Posts can be deleted, edited or quietly reframed, and nothing marks the gap afterwards. Pew Research Center found 18% of tweets vanish from public view within three months. Since losing calls are the ones most likely to disappear, the visible set always flatters the author.
8 min read
On the 2024 US presidential race the market leaned toward the eventual winner and the leading poll model did not. Polymarket had Trump at 58 percent on 4 November 2024, while the 538 model's final forecast gave Harris 50 in 100 and Trump 49 in 100. Pundits produced no scoreable number at all. One election does not settle the general question, and the same market data set looks much worse when you widen the sample.
6 min read
Money does not appear to buy accuracy. The head to head test that settled the question ran a real money exchange against a play money exchange across 208 NFL games in 2003 and found no statistically significant difference on four separate scoring rules. Cross platform Brier scores measured to October 2025 put the money and no money platforms in the same band. What money reliably buys is attention, liquidity and a way for large holders to push a price.
6 min read
A superforecaster is someone who ranked in the top 2 percent for accuracy across hundreds of scored questions and then kept doing it. The label came out of a research tournament, not a marketing department, and the advantage is measurable: better calibration, sharper separation of what happens from what does not, and a set of working habits that are cheap to copy. Nothing in the list requires talent you can only be born with.
9 min read
A prediction market price is a hard benchmark on average and a soft one in specific places. The price is most beatable when almost nobody is trading the question, when the information is not the kind traders watch, when the answer needs specialist knowledge, when the resolution rule says something different from the title, at the far ends of the probability scale, and where the price itself changes behaviour. Everywhere else, assume the price is right.
9 min read
