I spent time with the world’s top superforecasters

Last month at Kalshi’s inaugural Research Conference, I spent time with some of the world’s top superforecasters. Here’s what I learned.

The natural assumption one might make when qualifying the people that would be best at forecasting the future is that they come from a common background: high-achieving, Ivy League educated, finance professionals. This assumption couldn’t be further from the truth. 

There is no common pedigree. No shared industry. No obvious thread except that they are, by nature, extremely intellectually curious and, by choice, unbelievably knowledgeable about the world. In fact, I’d go so far as to say that the assortment of people I encountered was what some would term “random” - running the gamut from politics aficionados that are glued to the news and have mention markets down to a science, to people from towns across the United States who studied music, languages, or game design in college (if they attended at all). School teachers who happen to be excellent at tracking the Billboard Hot 100. Aspiring meteorologists who now closely follow weather markets. Poker players who reason about the underlying motivations of those around them. 

This finding, by the way, isn’t novel. Philip Tetlock, in his work with the Good Judgment Project (which in fact coined the term “superforecaster”), finds exactly this - that forecasting skill is largely decoupled from credentials. The best predictions of the future are rarely those whose job it is to predict the future, but rather those that are better, as Mellers et al. find, at “inductive reasoning, pattern detection, cognitive flexibility and open-mindedness”. 

Superforecasters attribute their success to disciplined practice rather than innate gift. They win because of what they do, not what they are. 

What does this look like in practice? Before engaging with the specifics of any question, they ask: what typically happens in situations like this? What drives the behaviour of the individuals or groups in question? They form a picture of the world driven by deep historical insights and views on human psychology, and update their perspectives as evidence contained in the enormous and anomalous quantities of news media consumed begs to change them. They have no attachment, emotional or otherwise, to their prior views. Their goal is not to be consistent. Their goal is to be right. 

None of this is exotic. It is just uncommonly disciplined. The forecasting skill, when examined manually, is largely an output of the process itself, rather than talent alone. This conclusion follows - approximately 70% of superforecasters tested by the Good Judgment Project retained top-tier performance year-over-year. This extraordinarily robust skill is why superforecasters I spoke to who trade high volumes manually - an action a number would view as nonsensical or archaic - are able to outperform algorithmic and high-frequency traders.  

The superforecasters that sit at the very top of the forecasting tree carry a deep, accumulated intuition for how the world works. 

They understand what drives and motivates individuals - power, legacy, money, pride. How institutions behave when they are under pressure. How and when political systems absorb shocks, and when they don’t. They have a sense, built up over iteration after iteration of worldview building, of how narratives form and collapse, how markets misprice tail risk, and how bureaucracies move. A calibrated view of what to look through as noise and what to ingest as signal. 

None of that lives solely in a numerical formula. Rather, it is the product of time spent paying very close attention to the world and critically assessing received observations. 

The quantitative scaffolding, still widely used, is what makes the intuition, in part, legible and accountable. But it is the intuition itself - the deep world-model - that separates the good from the excellent.  

One of the largest outstanding questions facing prediction markets today is the following - we know that our markets are well-calibrated, but why? 

Markets aggregate signals at speed and scale. But markets do not exist in the abstract. They are made of people - people who, at their best, look a lot like the superforecasters I spent time with just a few short weeks ago. Understanding not just what drives markets toward accuracy, but who, if anyone, is one of the most compelling open questions in forecasting today. 

A question that we, at Kalshi Research, are itching to answer. 

Find our research here: research.kalshi.com