How BrainPredict started: from big wallets to reading the price
At first we wanted to know why some large wallets always seemed to arrive too early. The answer took us elsewhere: less copying an address, more understanding why the price is wrong.
BrainPredict did not start as a marketing idea. We were watching Polymarket, and one question kept coming back: how do some wallets enter before everyone else, sometimes on thin markets, and exit at the right time?
We built tools to follow those public on-chain flows: entry size, timing, correlations across markets. It was instructive. We saw patterns. They were not always reproducible.
The anonymity wall
Transactions are public. Identities are not. Strong wallets change addresses, split orders, and blur the trail on purpose. Copying “the whale of the week” quickly becomes a game of whack-a-mole.
We stopped asking who was right and started asking why the market misprices an event. That change of question shifted everything.
Cross the facts, then decide
Rather than one model acting alone, or a permanent contest between AIs, we tested several architectures. Today the system crosses complementary sources: open news, visible large-wallet flows, structured market data, and, for geopolitics, context reading. A central synthesis then grades the file with a clear scorecard and filters before any public publication.
Several inputs, one decision
We ran several models in parallel. The current setup aggregates available facts, then decides with written rules. The goal is not to be right once by luck. It is honest calibration over time.
The public pipeline crosses news, market history, structured data and scoring before publication. For Polymarket access, start with the access and VPN guide. To read a price, continue with probabilities then odds formats.
Sign up free by email to receive analysis and PDF with every prediction. Optional SMS alert.