Numerai is a San Francisco hedge fund with a radically novel structure: every week, thousands of data scientists worldwide download anonymized, obfuscated stock market data, build machine learning models to predict price movements, and stake NMR tokens on their predictions — if correct, they earn NMR; if wrong, their NMR is “burned” (destroyed) as a Schelling point mechanism that incentivizes genuine skill over data snooping. Numerai then combines (“metamodels”) the predictions of its top performers to trade its actual hedge fund portfolio. NMR was one of the first cryptocurrency tokens with a built-in burn mechanism and skin-in-the-game incentives, making it a unique intersection of AI, finance, and crypto.
| Stat | Value |
|---|---|
| Ticker | NMR |
| Price | $8.43 |
| Market Cap | $59.40M |
| 24h Change | -0.4% |
| Circulating Supply | 7.04M NMR |
| Max Supply | 11.00M NMR |
| All-Time High | $93.15 |
| Contract (Ethereum) | 0x1776...6671 |
| Contract (Energi) | 0xd729...27a2 |
How It Works
Tournament mechanics:
- Numerai provides obfuscated financial data (feature-engineered, cannot be reverse-engineered to identify specific stocks)
- Data scientists build ML models and generate predictions for the upcoming week
- Models are submitted and stakers bet NMR on their model’s performance
- At the end of the round, stakers whose models correlate with actual market returns earn NMR; stakers whose models fail lose a portion of stake
- Top-performing models feed into Numerai’s live trading signals
Burn mechanism:
NMR is burned (not redistributed) from losers — it is permanently removed from supply. This makes NMR deflationary over time as tournaments run continuously.
Numerai Signals:
A parallel tournament for incorporating external stock signals (using your own data sources, not just Numerai’s), staked with NMR.
Numerai Era system:
Predictions are evaluated across “Eras” — training data has era labels making it non-sequential, so models can’t simply look up historical outcomes.
Tokenomics
| Metric | Value |
|---|---|
| Max Supply | 11,000,000 NMR |
| Deflationary | NMR permanently burned from losing stakes |
| Circulating | ~6.8M NMR (rest held by Numerai) |
| Tournament | NMR paid to winners each week |
| Foundation hold | Numerai holds ~4M NMR for future payouts |
Use Cases
- Tournament staking — NMR staked on ML model predictions in Numerai tournament
- Signals staking — NMR staked on external market signals in Numerai Signals
- Hedging — NMR held as exposure to an AI/ML-native crypto asset
- Burn mechanism — Losing predictions permanently burn NMR from circulating supply
History
- 2015 — Numerai founded by Richard Craib in San Francisco
- Jun 2017 — NMR token launches; Numerai airdropped NMR to participating data scientists
- 2018 — Erasure Protocol launches (decentralized prediction markets using NMR)
- 2019 — Erasure Quant and signals products develop
- 2020 — NMR ATH ~$160; significant DeFi and AI crossover interest
- 2021 — Numerai Signals goes live; tournament participants exceed 10,000
- 2022–2024 — Steady operation; tournament participants continue growing; NMR burn accumulates
Common Misconceptions
“Numerai predictions are real stock market advice.” The stock data Numerai provides is fully obfuscated — scientists cannot know which companies they are predicting for. The tournament tests pure signal extraction, not domain knowledge.
“NMR is just another hedge fund token.” NMR’s burn mechanism and tournament payout system represent genuine on-chain incentive engineering — it functions mechanically in a way most “fund tokens” do not.