EN
Back to the archive

The encyclopedia · R&D & Science · Technical decision · 2015–2018

Numerai encrypted its data so 40,000 freelancers could build one hedge fund's models

A hedge fund that cannot show its data runs a weekly tournament on encrypted numbers and blends the winners' models into one portfolio.

Numerai

the move

Numerai founder Richard Craib built his quant hedge fund on the idea that 'if one data scientist is good, 40,000 are better'. San Francisco-based Numerai crowdsources its strategy from thousands of data scientists worldwide who compete in a weekly tournament, using the firm's encrypted financial data to develop predictions about markets.

Encryption is the trick: outsiders never see the actual securities or prices, only anonymized features, so Numerai can open its data without leaking its edge. Winners stake Numeraire, a token built on Ethereum, and are rewarded when their predictions prove useful; the firm blends the submitted models into a combined meta-model.

In 2018 Numerai launched Erasure, a marketplace where anyone can sell predictive signals or data to hedge funds on the same token rails, extending the gig-economy model from its own fund to the whole industry.

why it works

  • Encryption makes it safe to share the crown jewels.
  • A weekly tournament turns the world into a recruiting pool.
  • Token staking ties pay to verified performance.
  • No data scientist ever sees enough to copy the fund.
the payoffCrowdsource models on data nobody can readinspired

what transfers

When your data is the secret, share it encrypted: give outsiders scrambled features and pay only for verified predictions, turning secrecy into the reason the crowd works for you.

what came after

Numerai kept running its weekly tournament and token-staking system for years after launch, and the Numeraire token remained the payout and staking currency; the Erasure marketplace idea fed into later Numerai data products as the broader 'prediction marketplace' wave grew.

references

spotted an error? The archive wants to know.

same kind of clever