Summary of 78 minute video interview with Richard Craib, founder of Numerai - reading time 6 min

Dear reader,

I was fascinated by the idea of crowd sourcing intelligent minds to solve problems since two heads are better than one and thousands are a LOT better than one. That's why I watched the long interview with Richard and summarised it for those interested.

I created a google doc summary with bold formatting for the important points for easy reading :
https://docs.google.com/document/d/11prXh1w1k8goSV-inPGM36I_sxxPp_2BQw46owkl-OA/edit?usp=sharing

Identical content from that document without formatting is below:

Secret and abstract datasets obfuscated from finance which come in the form of a huge grid of numbers between zero and one are made available to participators ranging from Masters to PhD students to Professors involved in the fields of mathematics, physics and computer science. They then have one week to create models which make predictions based off of the datasets.

The participators do not know what the model represents as it merely an abstract mathematical problem and Numerai do not know which models were used to make the predictions, meaning both parties rely on each other to make money out of this process if it continues - both parties are “blind”.

Numerai then combine hundreds of models together to create a model which is far superior than any one model created by one person, effectively crowdsourcing quantitative solution problem solving skills. Then they trade 100 stocks based on all the predictions.

Quantitative trading gets edges on a wide variety of securities like stocks
Axe capital from the show “Billions” is more about gaining insider information to gain an edge (usually illegal) whereas Numerai rely on high frequency trading which offer speed edges on a substantial number of trades with transactions sometimes lasting as short as a nanosecond.

A lot of data started becoming available recently in the finance industry due to increased computer performance so programmers started analyzing patterns were in this data to come up with predictions of future outcomes and they are called quants.

Normal quants often use statistical regression. Richard is a quant but he came at it from a machine learning perspective when working for a 15 billion dollar hedge fund which he left to start Numerai.

There is some animosity between quants and traders because quants earn about a million dollars by creating models which traders use to make 5-10 million dollard a year.

Sentiment analysis - computers collect data based on what people are saying in forums/social media to see if they can come up with predictions.
Computers can predict sentiment better than they used to but what is missing is the trade expression : If people like Google, it could simply be an advertiser. If many people like Google, that does not necessarily mean that the stock of Google will go up.
You would need human intuition to make decisions based on the sentiment and we don’t have that technology yet.

Numerai is pronounced like and is an iteration of Samurai.The currency is called Numeraire.

One company which does something similar to Numerai, i.e. they incentivize people from around the world to solve problems together is http://brilliant.org/

Numerai has a fund but Richard is not allowed to talk about it’s value.

Concordance means that predictions are coming from one model. Concordance checks for abuse, i.e. models which are good in general but not good on new data

Numerai experimented with giving participators higher and higher % of the profits and currently the profits gained from the models used are at the highest they have ever been.

Participators do not have to disclose their identity, they are simply paid based on the quality of their models.

Here is a screenshot of how much money the top datascients made (if you zoom in you can see that the top data scientist has accumulated 11.600 USD so far)

The cryptocurrency of Numerai is named Numeraire and is tied to the payouts of the company. It is not part of the equity of the company. Richard was not at liberty to disclose what % of the payouts were distributed to the participators.

There is no way that Numeraire can find out who it is they are giving money to, they only know that for example some of them are teenagers from Russia or doctorate professors from India because these users choose make themselves publicly known.

When they started there was a small conspiracy that the data models were being used for something potentially very bad like terrorism.

Richard is not allowed to disclose how often the models are working, but he says they are working very frequently.

Due to Trump’s new tax policy changes, Numerai will spend substantially less tax.
Interviewer: “So you will pay half as much tax?”
Richard: “Yea and we can spend that money better than the government”

Richard’s vision is for their data scientists to manage all the money in the world
If Numerai had 10 billion dollars in management, they would likely charge a 1% management fee for which they get 100 million per year with or without performance.

“I think with 10 billiion (in management) we would about about 10 million per month to data scientists.The people who are on the top of the leaderboard might be making one million USD per month.”

Richard is from South Africa and went to Cornell in the US to study maths.

The venture capital raised on September 2016 was enough to keep them going for another 18 to 24 months.

Team

Norman Packard - Co-founder of Prediction Company
Howard Morgan - Co-founder of Renaissance Technologies
Geoff Bradway - VP of engineering, formerly Google DeepMind
Olaf Carlson-Wee - Founder of PolyChain Capital, formerly Coinbase
Peter Diamandis - founder of XPRIZE, Singularity University
Joey Krug - Founder of Augur, Thiel Fellow
Yunus Saatchi - Permutation Ventures, formerly Vicarious
Richard Craib - Founder of Numerai

Hope you benefited from this,
hooked2thechain