Jan 31, 2022 · 33m · mad
Fireside Chat: Richard Craib (Founder & CEO, Numerai) with Matt Turck (Partner, FirstMark)
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this fireside chat hosted by Matt Turck of FirstMark, Numerai Founder and CEO Richard Craib details how Numerai combines quantitative finance, obfuscated machine learning data, crowdsourced modeling, and cryptocurrency tokenomics to build a market-neutral hedge fund.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 20.9% of the talking time here. How this is scored →
Matt as informed peer 2.9 Guest teaching 4.7 Guest disagreement 1.1 Matt pushing back 1.1
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:00 Critique of Accredited Investor Rules
Richard expresses frustration at financial regulations, calling it sad that net worth is used as the sole metric for investor sophistication.
Hardest push from Matt ▶ 7:40 Defining Alternative DataMatt politely interjects mid-explanation to define alternative data (satellite traffic, social media sentiment) for listeners before allowing the guest to continue.
Biggest teaching moment ▶ 29:42 Quant Track Record Statistical SignificanceRichard educates the host on how statistical significance in hedge fund returns can be proven quickly using cumulative binomial distributions across 1,000 simultaneous stock positions.
Matt holds his own ▶ 7:40 Explaining Alternative Data ContextMatt displays domain knowledge by breaking down financial industry concepts like alternative data and satellite tracking without needing the guest to explain them.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Track Record and Market-Neutral Investment Strategy | 2 | 5 | 1 | 1 | Matt prompts Richard to explain Numerai's impressive performance and asks basic clarifying questions about market-neutral strategies. Richard provides a clear educational breakdown of market-neutral investing and downside protection during crashes. | |
| Obfuscated Data, Machine Learning, and Alternative Data | 4 | 5 | 1 | 1 | Matt demonstrates domain knowledge by defining alternative data for the audience. Richard explains obfuscation techniques and reframes why long historical data is far superior to short alternative datasets for machine learning generalization. | |
| Participant Profiles, Learning Curve, and Numerai Signals | 3 | 4 | 1 | 1 | Matt asks engaging questions about user demographics and whether students can learn on the platform. Richard details user profiles from NASA and CERN and outlines Numerai's overarching master plan. | |
| Crypto Staking, NMR Tokens, and Meta-Model Weighting | 3 | 4 | 1 | 1 | Matt asks Richard to define cryptocurrency staking for uninitiated listeners and summarizes the core mechanism neatly. Richard explains NMR token mechanics, smart contracts, and burning bad stakes. | |
| Ensembling, Portfolio Optimization, and Staking Rewards | 3 | 5 | 1 | 1 | Matt inquires about ensemble techniques and user earnings. Richard explains why stake-weighted averaging outperforms complex dynamic weighting due to user confidence calibration. | |
| Revenue Model and Intellectual Property Security | 2 | 3 | 1 | 1 | Matt relays an audience question regarding Numerai's business model and intellectual property rights. Richard clarifies the traditional 2-and-20 hedge fund fee model and reassures that users retain full ownership of their code. | |
| Feature Engineering Edge and Crowdsourcing Signals | 3 | 5 | 1 | 1 | Matt relays a technical community question on feature engineering versus model architecture. Richard explains the incremental statistical gains that crowdsourced models achieve over baseline internal models. | |
| Asset Relationships and Regulatory Barriers to Retail Access | 3 | 5 | 2 | 2 | Matt brings up questions on individual asset modeling and retail participation. Richard offers a mild critique of regulatory accredited investor rules that prevent non-wealthy individuals from investing in hedge funds. | |
| Evaluating Quantitative Track Records and Final Remarks | 3 | 6 | 1 | 1 | Matt asks how long a quant track record takes to prove out before wrapping up the interview. Richard gives an in-depth explanation of statistical significance using coin-flip analogies and cumulative binomial distributions across multi-asset cross-sections. |
Statements from this episode (19)
Craib: Numerai has more data scientists than any other hedge fund
“And so we have the best modeling talent because we have much more More, many more data scientists working on Numeri than any other hedge fund in the world.”
Numerai fell only 1.5% during the March 2020 market crash
“And you know, for example, in March, 2020, when the market fell 30% in about 20 days we were down one and a half percent.”
Craib: Numerai spends over $1M per year buying data
“We probably spend more than a million dollars a year on data.”
Craib: A top Numerai participant works at NASA's Jet Propulsion Lab
“We have one really good user who works at NASA Jet Propulsion Lab and who's really strong at data science. And he's actually working on a mission to one of Jupiter's moons.”
Craib: Numerai's goal is to monopolize data and money, then decentralize it
“Numeri's mission is to kind of monopolize intelligence and monopolize data, and then monopolize money, and then decentralize the monopoly.”
Numerai builds meta-model using stake-weighted average of 5,000 global equity signals
“So everyone's providing signals on all 5000 stocks in global equities, and we take the stake weighted average of all those signals.”
Numerai rewards accurate models with NMR tokens and burns failing stakes
“And if your model performs well on Nimri, we'll give you more NMR tokens. And if your model performs badly, we can burn your stake.”
Craib: Numerai paid $10 million to data scientists in 2021
“We've even, I think even just last year we paid ten million dollars. So it's so much higher than the rewards are so much higher than the other data science competitions on online by orders of magnitude.”
Craib: Some individual Numerai contributors have earned over $1 million
“And there's some users that have made more than a million dollars.”
Craib: Numerai currently has over $15 million in crypto staked
“That's why I think we have, yeah, about right now over fifteen million dollars at stake.”
Craib: Numerai charges institutional investors standard 2 and 20 fees
“A big investor could come and give us a hundred million dollars and we would charge two and 20 on the investment.”
Craib: Traditional quant funds frequently collapse after one bad year
“What often happens with these quant funds is they have a good year, and then they have a bad year, and then they kind of die.”
Craib: Numerai users retain all source code and intellectual property
“They never give us their code. They never assigning us IP around their models. They can train them in any way they want to keep them on any server they want. The only thing they have to do is provide us the predictions every week.”
Craib: Numerai's crowdsourced meta-model vastly outperforms its internal ML models
“If we do say a linear model just on our data, it sort of performs okay. If we do our own internal machine learning model on the data, it performs better than that. And if we use everyone's model that's being submitted to Numeri, it's way better than even that.”
Craib: A 0.5% accuracy gain in quantitative finance dramatically transforms Sharpe ratio
“Especially in the finance domain where, let's say you're currently 52% right. If you can go to 52 and a half percent right, It's like a whole different level of performance for the end investor, a whole different level of shock ratio or returns and risk.”
Craib: Numerai hedge fund is closed to all but roughly seven investors
“At the end of this process, we have a hedge fund that is basically closed to everybody except for like seven investors in the world.”
Craib: Token staking serves to measure data scientist confidence in model generalization
“The whole point of staking is to get an assessment of how much they believe their model will Generalize.”