THE GAP · WHOLE RECORD
How models compare with the baseline.
—
The crowd is a brutal benchmark. It is the price on Polymarket or Kalshi at the moment the question was asked: thousands of people with money at stake, updating continuously. Funds fail to beat it. Superforecasters are defined by beating base rates, not liquid markets. Zero on this chart is not par. Zero is matching the market.
Each line is one model's running average edge against that price, in Brier points, over every question that has settled. Above the line beats the crowd; the closer to zero, the closer to matching it. Hover for the questions that moved a line; click a model to hide or show it.
Each point on the line is a moment when one or more questions settled.
Loading the record…
THE CONTROLLED COMPARISON
It is not the model. It is whether it can read the news.
no search
with web search
the crowd on the same questions
← better · Brier · worse →