Introducing Erasure Quant

Erasure’s first finance tournament

Max Novendstern
Jun 3 · 6 min read

Numerai built a stock-picking tournament on Erasure. We’ll use it to crowdsource alpha for our hedge fund. So tool up to make money. Quant is only the first in an ecosystem of tournaments on Erasure. So fork it and build your own.

We’re excited to announce Erasure Quant, the first dapp on the Erasure protocol. Erasure Quant is a stock-picking tournament that achieves something previously impossible: provable point-in-time stock predictions. Submit encrypted predictions to the blockchain. Reveal them to build reputation. Stake money on your next ones and sell to the highest bidder.

Quant demonstrates Erasure’s power as a finance tool. We’re solving a real problem — our own. We’re hunting for diversifying signals on stocks we trade, and want to enrich our community of data scientists along the way. But Erasure isn’t only for us. Quant is the first of many tournaments to be built by others on the Erasure protocol, towards a radically more open hedge fund industry.

How Quant Works

“A great vampire squid…jamming its blood funnel into anything that smells like money.” — Matt Taibbi on Goldman Sachs

Erasure Quant takes Numerai’s core insight — that markets beat hierarchies when sourcing signals about the future — and goes further. We open the market. On Quant, use any data you want. Own your reputation. Sell to anyone.

We’re launching Quant soon, along with the Erasure protocol. We’re announcing it now so you have time to tool up. Here’s what you need to know to make money on Quant. Email us at quant@numer.ai with questions.

Quant is an opinionated interface into the open Erasure protocol. With a few clicks in your browser you can encrypt predictions, upload them to IPFS, and send the hash to Ethereum, so they’re permanently timestamped.

Reveal your predictions so our “oracle” can score them against market data. We’ll display our reputation measures on your profile page, which anyone can verify independently.

Compete against a global community. We’ve organized the leaderboard around the metrics Numerai cares about, and we’ll continuously analyze the community’s performance, relative to the needs of our portfolio. We’ll bid for signals we’re attracted to.

Predict the outperformance and underperformance of stocks in the Russell 3000 index, for periods greater than one day. This CSV gives the total universe of company tickers to predict on and demonstrates the valid format.

  • Format: submit predictions as a CSV, with tickers on one column and signals between zero and one on the other. High scores mean outperformance. Low scores mean underperformance.
  • Size: the best signals predict the out/underperformance of a basket of two hundred-plus stocks. This makes our “Top-100, Bottom-100 Returns” measure (see below) meaningful.
  • Horizon: predict periods greater than a day. Your prediction CSV will be scored from the point you submit it to the submission of the next one. Your score for a given day is from the stock prices at the market open after your prediction was submitted to the prices at the next market open. (For example, if you commit a prediction CSV on Monday at 8:30AM EST and then commit another on Wednesday at 10:30AM EST, your first prediction CSV would be scored over three days, Monday through Wednesday.)

If you’re new to quantitative trading, we recommend Alpaca. Here’s an example of how to generate Quant predictions using Alpaca’s scripting language.

Once you reveal your predictions on-chain, we’ll score them against Quandl’s Adjusted End of Day US Stock Prices feed using two metrics: “Accuracy” and “Top-100, Bottom-100 Returns.”

  • Accuracy: for any given day, what percentage of the stocks that you predicted to be “outperformers” (top 50th-percentile of your scores) were in fact outperformers (in the top 50th-percentile of returns)? Vice versa “underperformers”? Your accuracy score for a day is the percentage of accurate predictions.
  • Top-100, Bottom-100 Returns: for any given day, how did a portfolio of the stocks you rated in the top hundred do? How about those in the bottom hundred? What’s the returns of a portfolio of your top hundred stocks divided by your bottom hundred stocks? Those daily returns will be posted. We’ll show the cumulative returns of your feed, the volatility of those returns, and then your compounded daily returns (like CAGR, but for days).

Build Your Own

Quant is a dapp. It is an example, we believe, of a better way to build internet applications. We have no privileged access to data, so anyone can re-implement Quant with a new skin. Data is as permanent as Ethereum and IPFS, so if we ever shut down you’ll still have all your contributions. The main workflows are run by smart contracts, so they’re transparent and governed.

Last week we announced the Numerai Grants Program. We put up $1 million in NMR for developers building on Erasure. In that post, we listed some tournaments we’d love to help you develop:

  • One for Tesla, calling for predictions across its financial statements
  • One for a frontier country stock market, like Kenya’s
  • One for crypto, both quantitative and discretionary

Quant is Erasure’s first tournament. But it’s just the beginning.

To roll your own tournament, you need four building blocks:

  1. Erasure Interface using Web3 or EthersJS (for example) to push data to the Erasure protocol and IPFS
  2. Erasure Querying Engine using The Graph (for example) to pull data from the Erasure protocol
  3. Reputation Oracle using data APIs and basic data engineering to calculate prediction accuracy
  4. Website for UX

We’ll release parts of the Erasure Interface and Querying Engine as packages, so they’re easy to work with. The rest is standard web development. Build your oracle, set up a site, and then make the market. Contact us at grants@numer.ai to learn about getting a grant towards building a tournament.

Erasure for the Long Run

“All the capitalists are worried about is where’s the productivity growth.”Marc Andreessen

It’s the age of low returns. Stocks are priced to be negative for ten years, and bonds to barely beat inflation. Warren Buffett counsels, at length, to buy index funds. But the agents of growth in the 21st Century— California’s tech startups and China’s urban developers; Redmond’s deep tech companies and Lagos’ engineering schools — are not in your 401k’s index strategy. Traditional financial infrastructure serves the winners of the past. You need to go to the frontier.

An open ecosystem of Numerai-style tournaments, for assets around the world — where a teenager in Somaliland or an engineer at a quantum compute lab can leverage provable reputation to sell information to a hedge fund in New York City––is a plausible answer to the question of how to build those “maps” of the frontier, how to reorganize financial infrastructure to better reflect the “territories” of today’s economic growth.

But don’t believe us. The truth of crowdsourcing’s value will be settled when so much money is made it cannot be ignored. So tool up for Quant. Or start your own. The truth, proxied by money, will reveal itself.

Numerai

A new kind of hedge fund built by a network of data scientists.

Thanks to Anson Chu, Richard Craib, and Natasha-Jade

Max Novendstern

Written by

Numerai

Numerai

A new kind of hedge fund built by a network of data scientists.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade