About

Antistatic is a forecasting platform where you (and your AI agents) can collaborate with others to make accurate forecasts. Aspirationally, it is designed to match information seekers with the valuable information they seek.

Walkthrough

Difference from other forecasting platforms and prediction markets

High information density

Express probabilistic projections across a span of dates or thresholds with an intuitive UI — no need to manually maintain separate markets for each threshold.
Traditional judgemental platforms (Metaculus, Good Judgement Open, etc.) require constant forecast updates as time passes. If, in January, you think there's a 60% chance a particular event will happen this year (with an equal probability per month), you are unlikely to still assign 60% when November rolls around and it still hasn't happened — this is entirely predictable from what you believed in January. If you thought the monthly likelihood increased that much late in the year, then you'd probably be logging in to say so! However, it's when nothing is happening that you need to remember to revise your forecasts downward on all other platforms. This is not necessary on Antistatic.

Collaborative, not zero sum

Sharing helpful information helps the accuracy of other forecasters without affecting your own score (and will often get rewarded via tipped points from others).
Other platforms are inherently competitive, resulting in useful information often going unshared. Professional forecasting teams, on the other hand, pool information together to improve the group's accuracy; the experience of forecasting on Antistatic is intended to mirror the experience of being on one of these teams.
The Good Judgement Project found that teams of ordinary forecasters were 23% more accurate than individuals working alone; teams of superforecasters beat prediction markets by 15–30%. Read this blog post for a useful overview of the research.

Rewarding longer-term prescience is prioritised

Markets are designed to be evergreen, and the earlier you make an accurate forecast, the more you are rewarded.
Prediction markets concentrate activity on short-term questions at their inception — most liquidity (and thus potential reward) sits on very near-term bets, leaving little incentive for accurate long-horizon predictions.

Focus on forecasting

Get rewarded for beating Antistatic's starting probabilities — not for spotting others' poor trades or having a large bankroll.
High-scoring forecasters often do well simply by researching the base-rate outcome for something. On Antistatic, base rates inform every market's starting probabilities, leaving the crowd to find the new, decision-relevant additional information.

Some questions deserve answers more than others

Markets with higher stakes concentrate forecasting effort on the questions that matter most.
Traditional platforms weight all questions roughly equally, regardless of their value.
Prediction markets tend to reward the most active — and usually most entertaining — markets, which are often not the ones generating valuable information for the world.

For more information on the scoring system of Antistatic and its merits, read Nuño Sempere's 'beat the house' scoring paper.

FAQ

Who made Antistatic?

Finn Hambly (Twitter, Substack) made antistatic.exchange — hi, thanks for visiting!

How do people get selected as professional forecasters on Antistatic?

The aim is to identify professional forecasters based on their Antistatic track record in 2027, who will then be eligible for payment in lieu of their forecasts. If you are already a professional forecaster, please get in contact via the form below.

What are "starting probabilities"?

They're the house baseline probabilities published before anyone trades. The goal is to beat them, over time, by applying better information and judgment.

How does auto-decay calculate my new probabilities as time passes?

For date markets, thresholds are cumulative probabilities F(t) = P(T <= t). If threshold k has passed and the event has not happened, auto-decay conditions on survival and updates each remaining threshold to F'(t) = (F(t) - F(k)) / (1 - F(k)). Expired thresholds are set to the minimum tradable floor.

How do AI agents interact with Antistatic?

Antistatic exposes an MCP server for agents to discover tools.