Numerai
Numerai is an AI-run, crowdsourced hedge fund based in San Francisco. It was founded by South African technologist Richard Craib in October 2015. Numerai’s trades are determined by Artificial Intelligence, which is fueled by a network of multiple anonymous data scientists. Numeraire (NMR) is an ERC-20 token that powers Numerai [1][2][30]
Overview
Numerai organizes a guessing game where people, often data scientists, create models to predict how money-related things like stocks will change. These models compete with each other. Those who make the best predictions win prizes. They can stake Numeraire Tokens (NMR) to win more if they are confident in their predictions. [22]
Richard Craib says in one of his Medium posts:
"The reason we ask data scientists to risk NMR by staking it on their models is so that only the good models get staked."[23]
Numerai won a 2016 Forbes Fintech 50 award. In March 2017, Numerai designed and released a new ‘API’ that allows people around the world to create and submit predictions from machine language models to power hedge funds. [6]
Numerai Categories
Tournament
Numerai hosts a tournament where data scientists are given an encrypted dataset weekly. This dataset is an abstract representation of stock market information and preserves the structure while keeping the data secret. The data scientists then use this dataset to create machine learning algorithms that find specific patterns. They can test their models by submitting their predictions to the Numerai website. [7]
If data scientists' models perform well, they earn more money (paid in Ether). If their models perform poorly, their Numeraire(NMR) is destroyed. [8][9]
Signals
Numerai combines the models submitted by data scientists into a meta-model that guides the hedge fund's trading decisions in financial markets. The platform uses the collective intelligence of the data scientist community to inform its trading strategies. [15]
Numerai Metamodel
The concept of a performant metamodel is predicated on the idea that a broad and diverse selection of uncorrelated models will more accurately predict the outcome of a complex data science problem. Particularly, in the case of predicting financial markets where high accuracy and low variance matter more. Numerai compares the input signal against all other signals to compute the given input signal's originality. [16]
Richard Craib provided a concise explanation of the principles, at NumerCon 2023, upon which the Metamodel operates. [27]
- CORR- The primary scoring metric on Numerai is the correlation between the model's predictions and the target. Numerai uses a variation of correlation- Numerai Corr. At a high level, this metric is designed to be a good proxy if the predictions were used in live trading.[25]
- FNC- Feature neutral correlation (FNC) is the correlation of a model with the target after its predictions have been neutralized to Numerai's features. [26]
- TC- The above two evaluation indicators were not successful in reaching their ideas. So, Numerai has invented a signal evaluation index method called True Contribution. True Contribution is computed by treating Numerai as an end-to-end artificial intelligence system. [24]
The concept of TC (True Contribution) is that Numerai participants submit their predictions, stake NMR if they are confident, an ensemble is performed using the weighted average, a portfolio is constructed by the optimizer, and the fund manages the long-short strategy to generate profits. The series of Numerai flows that are created are likened to one neural network (NN). [27]
Technology
Erasure Protocol
In October 2018, Numerai introduced the Erasure Protocol. The Erasure protocol is a set of smart contracts on the Ethereum blockchain that enables Numerai to either reward users who have staked their NMR coins to a correct outcome or destroy it if they were wrong. The protocol tracks and records the history of predictions, proving that the user had the success rate they claim to have and that any information a user wants to sell isn’t edited by anyone. [30]
The idea is to create a self-policing marketplace where the most honest get rewarded, and the dishonest, lose their staked crypto. [12]
The two Numerai platforms that use Erasure include:
- Erasure Quant - a dApp that crowdsources predictions.
- ErasureBay - an open marketplace for any kind of information.
Dataset
Numeraire’s dataset is a collection of regularized financial data that is used in the Numerai tournament, a data science competition where participants build machine learning models to predict the stock market. The dataset is obfuscated so that it can be given out for free and modeled without any financial domain knowledge. It is assembled from several market data providers and the traditional and unique features are consistently researched. [16]
The latest version of the dataset is V4.3 which was released in January 2024. It contains 244 new features with a total of 2376. [32]
Products
Numerai One
This is Numerai's flagship market-neutral fund. It seeks to deliver consistent returns in any market environment with a highly diversified portfolio that is market, country, sector, and factor-neutral. [28]
Numerai Supreme
Built on the same investment framework as Numerai One, Numerai Supreme delivers higher returns with higher volatility. This is achieved through a concentrated portfolio of the Meta Model's highest conviction selections while remaining market-neutral. Numerai Supreme was launched in August 2022. [27][28]
NMR Token
NMR is a digital asset token used to pay for services on the Numeraire network. It is built on Ethereum following the ERC-20 standard for tokens. Numeraire can be bought and sold for other digital currencies. [10][11]
In 2017, the NMR token launched on Ethereum mainnet without an Initial Coin Offering (ICO), rather Numerai issued one million NMR tokens to 12,000 data scientists based on past contributions in Numerai Tournaments. [30]
The maximum supply of NMR is 11,000,000 tokens. In a 2018 press release, Numerai announced that three million NMR tokens will remain locked until 2028, and will be released as a reward incentive for tournament participants. NMR tokens are burned weekly, as tournament participants who venture losing models and predictions lose their stake in the competition. [21]
Funding
On April 6, 2016, Numerai closed a $6 million Series A funding round led by Union Square Ventures. On December 12, 2016, it received $1.5 million in Seed funding, led by Howard L. Morgan, founder of hedge fund Renaissance Technologies, and Naval Ravikant. [5]
In March 2019, Numerai raised $11 million from a token sale led by Paradigm, Placeholder VC, and others. In June 2020, Numerai closed a $3 million NMR token sale for Erasure expansion with participation from Union Square Ventures, Placeholder VC, and others. On February 2, 2023, Numeraire raised $10 million in a Series B round led by Richard Craib, the CEO of Numerai, FJ Labs, Shine Capital, and Union Square Ventures. [29][13][31]
Team
The Numerai team includes:
- Founder & CEO | Richard Craib
- CTO | Anson Chu
- COO | Dan Spier
- MD | Rich Allen
- Director of Portfolio Management | Rui Tang
- Portfolio Manager & Fundamental Data Analyst | Marc Beckenstrater
- Chief Data Officer | Michael Phillips
- Chief Science Officer | Michael Oliver
- Director of Operations | Dennis X.
- Software Engineer | Bryant Luu
- Data Scientist | Ubajaka Chijioke[14]
