Numerai: Performance sharing of a quantitative hedge fund established by global data scientists

Author: Numerai

Compiled by: ChainCatcher

Numerai is a quantitative hedge fund built by thousands of data scientists around the world. We have quietly outperformed some of the most prestigious hedge funds in the industry, but we have never disclosed our performance. Today, we are excited to share the performance of our hedge fund for the first time with the community of data scientists who built it. This blog post was updated on February 3, 2022, to reflect Numerai's performance as of December 31, 2021, and to update the fee methodology used to calculate net performance.

We launched the current version of the Numerai hedge fund in September 2019. Around this time, we changed the dataset, optimizer, leverage, and the way the Numerai tournament operates (see Implementing Meta Models First). This was an incredibly challenging time to begin establishing a performance record for a quantitative hedge fund. Within months, Covid-19 triggered massive volatility in global markets, leading many quantitative hedge funds, including Renaissance's RIDA fund, to record their worst periods ever, with the fund down -32% by the end of 2020.

However, Numerai's market-neutral hedge fund performed exceptionally well. Despite the challenges of 2020, it allowed Numerai to stand out among blue-chip quantitative hedge funds. It highlighted the risks faced by some of those older funds. Many of these funds had significant factor exposures that Numerai's fund avoided.

Since the strategy's launch in September 2019, Numerai has outperformed Aurum's Quant Equity Market Neutral Index by 26.46% and AQR's Market Neutral Fund by 29.61%***. The past two years have been quite good.

The correlation of Numerai's return stream with AQR is also very low, at just 0.01, indicating that Numerai is indeed doing something completely different from traditional quantitative analysis.

Market-neutral hedge funds are different from any other investment product. Most funds have agreements with investors to take on risk (their "authorization"). Venture capitalists are responsible for deploying funds into tech startups. Cathie Wood is tasked with taking risks on innovative stocks. On the other hand, market-neutral funds require investors to remain neutral to stock market risk, meaning they hold portfolios that do not contain that risk. It is this authorization that allows them to run portfolios with very low risk, which may lead to returns that are much lower than when the market is up, but with relatively much higher returns during market crashes.

The key idea of a market-neutral fund is that for every $100 of stock bought, $100 of stock is shorted. If this condition is maintained, the net effect is that the hedge fund's market risk is approximately $0. Roughly speaking, this is what it means for a portfolio to have a beta of zero, which is how market-neutral funds operate.

Of course, many portfolios with a beta of 0 still carry risks. For example, if a hedge fund buys $100 of U.S. stocks and shorts $100 of Japanese stocks, the portfolio will face significant country risk. (In particular, the risk that U.S. stocks underperform Japanese stocks). Buying $100 of Apple and selling $100 of Microsoft might be a safer portfolio since both stocks are at least in the same country and industry. This long position in Apple and short position in Microsoft would be considered country and industry neutral.

Hedge funds like Numerai are trying to remain neutral to market, country, industry, and hundreds of other risk factors. The greater the neutral risk of the portfolio, the lower the volatility of the portfolio, and the more we believe that the risks of using leverage will be reduced.

So yes, it is confusing to read about a hedge fund with 50 PhDs making only 5% in a year when everything is up, but when you consider how small the risks taken by the hedge fund are and how low the volatility is, 5% can be impressive. A 5% return is much higher than what many corporate bonds pay, and the risks and volatility of the hedge fund may also be lower than those of bonds.

Good question. If hedge funds can make money with a beta of 0 (we must short every time we go long), this almost always means that the stocks we are long on will outperform the market.

One reason Numerai started as a market-neutral fund is precisely because it is the hardest game on Wall Street. It is difficult because the only way to make money is to be good at stock selection and risk management. With the market-neutral authorization, taking on market risk does not earn money for free. So we believe that Numerai data scientists can win in Wall Street's hard mode, and then we can build any type of fund in the future. It took many years, but we are now excelling in Hard Mode, and due to Numerai's rapid growth, we are getting better at an accelerated pace, which is very rare for a hedge fund business model.

So if you ignore all the shorts and build a portfolio solely from our long positions, this is how Numerai has performed.

9/9/2019--10/01/2021 Numerai US Long Only: +98.25% iShares Russell 2000: +51.93% Numerai Outperformance: +46.32%

Numerai's U.S. long portfolio returned 98.25%, while the Russell 2000 returned 51.93%. Please note that this is not backtested or simulated. According to our audit records, these are the actual stocks Numerai purchased at the prices we actually paid. Of course, the funds above are not real funds (our real fund has many offsetting shorts and leverage), but this is just an example of how Numerai's U.S. longs compare to the stock market benchmarks that most people are familiar with. We chose the Russell 2000 instead of the S&P 500 because the stocks in the Russell 2000 are smaller in size and more closely match Numerai's U.S. long holdings. (We also beat the S&P 500 index, which "only" rose about +43% during this period, while Numerai was at 98.25%.)

These are not definitive evidence that Numerai's fund performance is driven by skill rather than luck, but they are enlightening. So is this:

If a fund is lucky, it typically performs well in the long run due to its risk exposure, while performing poorly in the short run. Compared to the Russell 2000, Numerai seems to have a strong and consistent relative outperformance, which will become stronger as the number of data scientists staking models on Numerai increases. Numerai has outperformed the index on 54.36% of trading days, which is hard to achieve by luck over 539 trading days.

The tracking error of the portfolio is 7% (for non-quants, this is a measure of how much the portfolio varies from the benchmark). This indicates a risk of 7% relative to the benchmark, which we consider quite low. The information ratio is 2.08 (for non-quants, this is somewhat like a Sharpe ratio for returns above the benchmark).

These are indeed very good numbers. I think they would place Numerai at the top of the U.S. long fund category. There are not many funds with an information ratio of 2.08 (0.5 is considered a good information ratio). So, congratulations to Numerai data scientists around the world whose models have created this outstanding record.

Diversification. The S&P 500 does not usually do this as recently. Sometimes the market crashes and stays down for a long time. From 2000 to 2010, you could hardly make any money holding stocks in the U.S. Another similar decade is ahead of us, and when it arrives, it will be wise to hold something that can survive and thrive, like a market-neutral hedge fund.

Even if institutional investors expect Bitcoin to outperform the stock market, it does not mean they should put all their money or even most of their money into Bitcoin. Institutional investors care about things like volatility-adjusted returns because they are in a game that spans decades, where they will see multiple environments. Because of their lack of correlation with the market and more risks, market-neutral funds will always have a place in the portfolios of sophisticated investors. They may even be the perfect diversifiers in the world, having extremely high valuations for every risk asset.

Moreover, almost any institutional investor can acquire stock exposure, tech exposure, or Bitcoin exposure for free. These are just risks, so the cost of creating and managing them is low. It is difficult to create a market-neutral fund, and therefore it is also difficult to replicate it at a low cost. Institutional investors are often willing to pay fees to market-neutral hedge funds like Numerai for unique, uncorrelated diversification risk exposures that they cannot obtain elsewhere.

We recently raised funds from a new investor. They are Numerai's first large institutional investor, a pension fund. They reached an agreement with Numerai to give us $20 million, with the option to reinvest $130 million later. The follow-on investment after the initial $20 million has a lower fee, so they are incentivized to use their $130 million capacity rights to reinvest when new investors come on board.

Our current assets under management are only about $43 million, so we are still small for a hedge fund. But Numerai's fund trades with a leverage of 5.5 times, so nearly a quarter of almost $1 billion is controlled by Numerai's data scientists. Pretty cool.

It is only a good idea to raise assets for a quantitative fund when it reaches some important milestones. For Numerai, we want to see four things before raising assets:

It took some time, but today we are at 4/4:

Learn more or contact us at numerai.fund

Numerai Master Fund One, Ltd., hereinafter referred to as "Numerai" This blog post was updated on February 3, 2021, to reflect Numerai's performance as of December 31, 2021, and to update the fee methodology used to calculate net performance. *** AQR composite return period references N class monthly returns ------ QMNIX. [†]: Numerai performance report net of fees using Founders Class structure: 1% management fee and 20% performance fee.

Numerai hedge fund performance results are for reference only. Numerai's fund performance results are calculated after deducting management fees and incentive allocations, assuming that management fees are paid in advance at a rate of 1% (1/12 of 1% per month) of the net asset at the beginning of each month, and annual incentive allocations at a rate of 20% of trading profits. The method of calculating the return rate is to divide the net performance by the net assets at the beginning of the period. Individual investor performance may vary due to different management fee and incentive allocation arrangements and the timing of contributions and withdrawals. Returns include reinvestment of dividends and other earnings, including income from new issuances. Returns may vary for investors restricted from participating in new issuances. Performance results have been reviewed and audited by independent accountants. The information provided is historical and not a guide to future performance. Investors should be aware that investment losses are possible. It does not represent that any investor will or may achieve profits or losses similar to those shown.

Aurum Quant Equity Market Neutral includes hedge fund information for the designated "quantitative equity market neutral" strategy published by Aurum Research Limited. Such information is unaudited and provided "as is."

Due to the tax treatment of short-term capital gains, Numerai's hedge fund is unlikely to be a suitable investment for individual investors. Numerai's hedge fund is primarily designed for institutional investors such as endowments and pension funds.

Numerai's cryptocurrency token NMR is the collateral token used by Numerai's data scientists on Numerai. NMR is a volatile cryptocurrency that fluctuates up and down with supply and demand along with other cryptocurrencies.