THE CROWD BELIEVES IN YOU, A LITTLE TOO MUCH

On Manifold, a prediction-market website, users routinely ask questions about themselves: “Will I go to the gym three times this week?”, “Will I finish my thesis draft by Friday?”, “Will I stay off social media in March?” Other users then trade on the answer with mana, the platform’s play money. Each question ends up with a price — read as the probability that the answer will be YES — which can later be checked against what actually happened.

Onlookers, the textbook cure for optimism

Psychology has a clear expectation here. People forecasting their own behaviour are optimistic: they underestimate how long their tasks will take and overestimate their own follow-through, a bias known as the planning fallacy. Observers judging the same plans lean on base rates — how often such plans usually succeed — and are less optimistic. A prediction market should add to that advantage, because it pools many observers and pays those who correct a wrong price. But a market about one person’s week is thin: few traders, little money at stake.

6,150 bets on “Will I…?”

Zihan Chen and Di Zhu, of the Stevens Institute of Technology in Hoboken, New Jersey, and Lei Nico Zheng, of the University of Massachusetts Boston, retrieved the list of all 194,347 markets created on Manifold from 2021 to 2026. They kept 6,150 resolved self-markets — forward-looking questions in the first person — opened by 1,660 people. Their benchmark: 6,272 external markets on outside events, such as elections, sports or technology releases, made by creators who never asked about themselves.

From the platform’s order records, the team rebuilt every market’s price path and every executed purchase. That told them who moved each price: the person whose future was at stake, a human observer, or an automated trading program. Two research assistants checked the reconstruction against the live website and, with their own code, recomputed 23 of the reported figures.

About four points too hopeful

A market is calibrated when its outcomes happen about as often as its prices say. External markets nearly were: they missed by 0.9 percentage point — a percentage point being the plain difference between two percentages. Self-markets were not. Their final prices averaged 49.4%, but they resolved YES only 45.0% of the time: an overstatement of 4.4 points. The gap is widest where the outcome was most uncertain. Among markets closing between 20% and 80%, it reached 9.4 points, and a self-market priced at one half came true only 39% of the time, against 47% for an external market at the same price.

Self-markets are smaller and shorter than external ones, but that does not explain the gap away. Compared at the same price level, with size and duration taken into account, the difference is still 2.2 points. And among the 1,222 people who created both kinds of market, their questions about themselves still ended up priced 1.9 points more optimistically than their questions about the world.

Born in the first day of trading

The person does not build this optimism into the opening price. At the start, 95% of self-markets sat at the platform’s default of 50%, and at the same opening price they actually came true 3.0 points more often than external markets. The overstatement appeared with trading: one day after creation, self-markets were already 2.4 points more optimistic than external markets at the same price, and the gap peaked at 5.3 points a week before close.

It arose even where the person never traded. In the first day, human observers sold external markets down by 4.5 points, from 50% toward the lower rate at which such events happen. Self-markets about goals came true less than half the time too — 44% — yet observers moved them only 0.8 point down, and where the person never traded, not down at all. Observers applied the outside view — judging a case by how similar cases usually turn out — to events, not to people.

One exploratory exception stands out. In the 294 markets whose YES was an outcome the person would not want — relapsing, breaking a streak, falling ill — observers sold prices down by 6.4 points in the first day. These markets were not overpriced: their final prices came within half a point of how often the misfortune actually happened.

The person holds the answer

Over a market’s whole life, trading moved self-market prices 27.8 points toward the eventual outcome on average, against 37.3 for external markets. The person’s own orders supplied 38% of that movement, against 9% for the creators of external markets, and nearly half of their share came in the final day before close. Their purchases also picked the winning side more often than strangers’, mostly near the end.

Familiarity did not help. Observers who had interacted with the person on the platform before picked the winning side no more often than strangers. Nor did the person’s track record: prices followed their record on earlier self-markets, but beyond the price, that record added no detectable predictive value.

What no formula can add

Since the bias is systematic, could prices simply be corrected? A correction learned from markets resolved before 2025 lowered the error of later self-market prices — measured by the Brier score, the mean squared distance between price and outcome — by only 1.5%, too little to tell apart from zero. The reason: beyond leaning toward YES, self-market prices separate future successes from failures less well than external ones, and no transformation of a price can supply that.

Why observers price people this way, the data cannot settle. They may share the person’s hopes, or simply read a stated goal as evidence of what will happen. The markets use play money, within one online community, and the analyses were not preregistered. As hypotheses still to be tested, the authors suggest that platforms showing a crowd’s view of someone’s goal could display how often similar goals by similar users succeeded, and invite the person to report progress or state their own probability. In their reading, the information the crowd lacks is held by the one person whose future is being priced.