A new ethereum-based token hit Bittrex yesterday (23rd June 2017). No, it didn’t do an ICO. And no, you shouldn’t dismiss it as “yet another token on the ethereum platform”. It’s called ‘Numeraire’ (https://numer.ai/) and this one demonstrates significant promise. Don’t take my word for it. I encourage you to read my findings and add your thoughts to the discussion. I will refrain from adding unnecessary commentary, and I will try my best to present the objective facts.
Before I invest in a coin, I write a fact-sheet that collates all publicly accessible and critical data of a project. I’ve decided to start posting my research. They’re quite detailed for the average post. If you find it helpful, please let me know. Additionally, if you have any tips or feedback, I would greatly appreciate it.
1. The Proposal
1.1. Key details about the project
In short, Numerai is a prediction-market platform. Numerai provides data that can be modeled with machine learning. The participants submit predictions based on this data and may stake NMR tokens to demonstrate 'confidence' in their prediction. A participant demonstrates confidence by the NMR staked per 1 dollar potentially earned. These predictions are used by Numerai to build its own metamodel and trade as a hedge fund. After 4 weeks, the performance of the predictions are evaluated. Those who performed 'well' (refer to the excerpt below) earn USD in the form of BTC (or soon ETH) from a prize pool. The prize is distributed in descending order from the participant with the highest NMR staked per 1 dollar until the pool is exhausted. Any participants who under-performed and were considered prior to exhaustion will “risk having their Numeraire destroyed,” which is irreversible and publicly verifiable on the Ethereum blockchain. After the prize pool is exhausted, no further participants are considered for distribution. These participants will have their NMR returned.
You can see the recent tournament results here: https://numer.ai/history/57
It is important to remember that the value of the token is not directly influenced by the monetary value produced by the hedge fund. There are no dividends for holders. The token is used as a stake and expression of confidence in a submitted prediction. If the participant is successful, they will 'win' a greater proportion of the prize money. Aside from speculation, the value of the token is contingent upon its demand by users wishing to participate in the tournament. It can be assumed, however, that the successes of the hedge fund will generate greater demand for the token.
Here is an excerpt from the whitepaper that explains the process in greater detail:
Every tournament has a staking prize pool, which is some fixed number of dollars. The auction mechanism allocates the prize pool among data scientists. Data scientists can submit bids to the auction. Bids are tuples (c,s)where c is confidence defined as the number of Numeraire the data scientist is willing to stake to win 1 dollar, and s is the amount of Numeraire being staked. For some time t, s is locked in the Ethereum contract, inaccessible to anyone, including Numerai. After t has passed, a variant on the multiunit Dutch auction is used to determine the payouts.
The auction mechanism is a multiunit Dutch auction with some additional rules. Performance is evaluated after time
t. The performance evaluation metric is logloss 2, a suitable metric for binary classification problems like Numerai’s machine learning competition. A model is considered to have performed well if logloss < - ln(0.5) , and badly if logloss ≥ - ln(0.5). The data scientists are ranked in descending order of confidence c. In descending order of confidence until the prize pool is depleted, data scientists are awarded s/c dollars if their models performed well or they lose stake s if they perform badly. Once the prize pool is depleted, data scientists no longer earn dollars or lose their stakes.
1.2. Key details about the token and its distribution:
• Built on the ethereum platform;
• 21 million max cap;
• 2.2 million total cap (as of posting this);
• 1,223,451 million was released to 'data scientists' (ie users of the platform);
• NMR is minted each week and distributed until the max cap is reached;
• NMR is used as a stake for submitted predictions; and
• NMR is destroyed for under-performing stakeholders.
2. Key People Involved
I must point out, again, that the Numerai team did not host an initial coin offering (ICO). Instead, they sought seed and series A funding, which totals an amount of ~7.5 million. In respect of this amount, the CEO, Richard Craib, stated to ETHNews:
“We don’t need to do a crowdsale, because we already have the money to make this work.”
2.1. Meet the team
Aside from funding, which will be addressed soon, there are a few notable names developing the project.
• Richard Craib (CEO): Richard worked in hedge funds and start-ups. He was announced as one Forbes 30 under 30 in Finance (https://www.forbes.com/pictures/ghmf45eeelg/richard-craib-29/#a294bb670f6d). There is an interesting article written by someone who knew Richard Craib when Richard was studying Mathematics at Cornell University.
“I went to school with Richard at Cornell. Richard studied math. Quiet and studious disposition. Played chess and poker. Traded stocks and started companies in his free time, learned to code for fun” – Francis Pedraza
• Geoffrey Bradway (Vice-President of Engineering): Geoffrey is a software engineer who, according to his Linkedin, helped develop Google’s Deepmind AI and “critical core [Youtube] infrastructure”. Geoffrey worked at Google for close to 2 years. He is now the Vice-President of engineering at Numerai. https://www.linkedin.com/in/geoffrey-bradway-68115683
• Xander Dunn (Software Engineer): Xander is a software engineer who, according to his LinkedIn, was a former developer for Apple Inc. and worked on bug fixes in iOS and Mac OSX (as well as various other projects). Xander worked at Apple for 1 year and 3 months. Xander was, also, a developer at Osaro, Inc, which is a start-up developing “deep reinforcement learning”. He is now a software engineer at Numerai. https://www.linkedin.com/in/xanderdunn/
• Norman Packard (Advisor): Norman is a chaos theory physicist who is renowned for his contributions to chaos theory and cellular automata. He, also, founded ‘Prediction Company’, which uses statistical learning and forecasting to build trading systems. (https://en.wikipedia.org/wiki/Norman_Packard )
2.2. The backers
There are some big names funding Numerai. I will list the notable backers. All information Is sourced from crunchbase (https://www.crunchbase.com/organization/numerai#/entity.)
Numerai raised 6 million in Series A funding (December 2016) and contributors included:
• Union Square Ventures (led by Andy Weissman) for $3 million. USV is headquartered in New York and is one of the biggest venture capital firms in America. It handles a portfolio of $1 billion in assets and was once rated the top VC in the world. Some of its prior investments include twitter and Zynga. https://www.businessinsider.com.au/its-official-union-square-ventures-is-the-top-vc-in-the-country-2011-4/?r=US&IR=T)
• Co-founder of Coinbase, Fred Ehrsam.
Numerai raised the remaining 1.5 million in seed funding and other investments.
• Co-Founder of Renaissance Technologies, Howard Morgan. Renaissance Technologies is a highly successful hedge fund that pioneered mathematical and statistical analyses for quantitative trading (similar to Numerai). It has produced $55 billion in profit over 28 years and is “about $10 billion more profitable than funds run by billionaires Ray Dalio and George Soros” (https://www.bloomberg.com/news/articles/2016-11-21/how-renaissance-s-medallion-fund-became-finance-s-blackest-box)
• Co-founder of AngelList, Naval Ravikant.
• CEO of Augur, Ron Bernstein. Augur is a prediction market built on Ethereum.
• CEO of Parse and Founder of Scribd, Tikhon Bernstam. Parse was acquired by Facebook for $85 million (https://techcrunch.com/2013/04/25/facebook-parse/ )
Special thanks to the NMR slack channel for kindly assisting with my understanding of the project.
Please Note: fact-sheets take a lot of time to create. Please let me know if you find them useful. Any feedback is appreciated!