A little-known virtual currency fund heavy on math and statistics is turning a steady profit under a lead trader who once oversaw a major cryptocurrency exchange.
French mathematician Évariste Galois, for whom Galois Capital is named. (Wikimedia Commons)
Galois Capital, a San Francisco cryptocurrency hedge fund that launched in January 2018, said in an investor letter and financial filings that it increased its holdings from $10 million to $23 million in two years with high-frequency trading and funds from new investors lured by returns. A quantitative fund manager, Galois Capital computationally makes bulk volumes of speedy, precise trades shaped by its founder and lead trader Kevin Zhou, previously head of trading at American cryptocurrency exchange Kraken, and a team of technical talent.
“Most of us are math, physics or computer science folks,” Zhou told CoinDesk, adding that highly sought International Mathematical Olympiad competitors have been sending in their resumes.
Quantitative approaches to trading are wide-ranging. They might encompass regression modeling, direction and magnitude calculations for price prediction or stochastic processes for volatility modeling and options pricing, Zhou said. On the technology side, programming interfaces, trading software and hardware equipment enabling fast communications and data analysis with exchanges are used.
Galois Capital built much of this architecture from the ground up, with custom tools such as co-located servers and network adapters, due to what Zhou cites as a dearth of heavy-duty options for cryptocurrency traders found at Hudson River Trading and Jane Street Capital, two of the biggest Wall Street quantitative trading funds.
Overall, the quantitative bent is still relatively tame at Galois Capital. For Zhou, sophisticated trading models and technologies, such as the machine learning software the fund experimented with and shelved, are sometimes overkill in this era of cryptocurrency markets.
“What works is a lot simpler than what would work in traditional markets,” said Zhou. “Some of these models that we have right now would not work in traditional markets, in more mature and more efficient markets.”
While trading is easier in the crypto-asset class than in traditional asset classes, Zhou said, the field has gotten more competitive since he managed from 2013 to 2017 at Buttercoin, a bygone bitcoin exchange backed by Silicon Valley-located startup incubators Google Ventures and Y Combinator, and from 2015 to 2017 at the trading desk at Kraken.
At Buttercoin, “the sizes of transactions were a lot smaller. The spreads were a lot bigger. I remember there were days where you were getting 100 bps [basis points] just trading $100,000,” said Zhou. “At Kraken, spreads tightened up a bit. It was like 40, 50 bps on $200,000. Now, it’s a lot tighter, probably a million dollars gets 10, 15 bps.” A spread is the difference between a financial instrument’s bid and ask price; a “bp” (pronounced “bip”), or a basis point, reflects a 0.01% change in a financial instrument’s value.
Considering Kraken is valued at $4 billion and processes millions of dollars in cryptocurrency flows each month, Kraken’s trading desk was an all-seeing looking glass into why cryptocurrencies are bought and sold in a large corner of the market. It gave Zhou a knack for sizing up counterparty motivations with Galois Capital’s programmed trades, where the other trading actor is faceless.
“When you’re market-making with bots on all these different exchanges, you don’t actually get matched up with on the other side,” Zhou said. At Kraken, his trading desk dealt with miners and investors, up close and personal, who gave context into their market movements. “So just being able to read Kraken’s book,” as in the exchange’s record of buy and sale orders, “is definitely informative.”
Market-making as a “safer” line of business
In terms of trading volume, Galois Capital went from processing $671 million to $1.4 billion between 2018 and 2019, the investor letter says. While non-algorithmic trades shrunk from $666 million to $562 million, its algorithmic trades blew up from $5 million to $876 million to enlarge the fund’s share in crypto-asset markets.
According to Zhou, about 85% of Galois operations are geared towards liquidity provision, matching crypto-assets at prices quoted by bidders, similar to services offered by Genesis Trading, Cumberland DRW and Circle. The other 15% of operations were focused on hedge fund management of cryptocurrency plays. Galois Capital started with over-the-counter (OTC) trading — manual liquidity provision and Zhou’s specialty at Kraken — and branched out into algorithmic market-making — automated liquidity provision — and discretionary trading.
WITH MARKET-MAKING, GENERALLY YOU DON’T WANT TO BE HOLDING ONTO RISK FOR THAT LONG. WHAT I MEAN BY THAT IS MORE THAN 30 SECONDS.
“Liquidity provision in the traditional markets is handled by prop shops rather than hedge funds. So we’re in kind of a unique situation in that most crypto funds are long only in these different tokens, or they’re long-short and look at factors like trying to detect momentum signals and reverse signals,” Zhou said.
“I want to be able to generate profits regardless if the market’s moving up or down, regardless if there’s momentum or reversion, just based on the micro-structure of the market, just based on providing compensation for providing liquidity to the market,” Zhou said. “To me, that seems a lot safer and generally, as a trader, I’m more conservative.”
Long-short, but for the short term
Galois’ hedge fund wing, which emphasizes derivatives, quantitative long-short and discretionary long-short trades, as of January 2020 netted to-date 29.5% for a Class A fund and 53.5% for a Class B fund. (The funds charge different fees on investment subscriptions, which begin at $50,000 an investor.) The two-year returns outdid several benchmarks over the corresponding period, such as a 20.8% S&P 500 gain, a 46.6% bitcoin loss, a 32.6% cryptocurrency hedge fund loss and a 4.6% non-cryptocurrency hedge fund gain, according to the CoinDesk Bitcoin Price Index and Eurekahedge return indices.
The Class A fund was, however, more volatile than the Class B fund: According to the investor letter, the Class A fund tumbled 22.4% in 2018 and spiked 66.8% in 2019; the Class B fund ticked up 33.9% in 2018 and 19.3% in 2019.
“We don’t really have that much of a long-term portfolio that holds the large positions longer. For us, a lot of this is just very short-term inventory balances,” Zhou said to CoinDesk. “With market-making, generally you don’t want to be holding onto risk for that long. What I mean by that is more than 30 seconds. So we’re usually flat on that exposure at any given time. And then, for very short periods of time, we’ll have short exposure in a number of different coins, but we’ll hedge that off very quickly.”
Galois Capital opened cryptocurrency futures and swaps in April 2019 and long-short trades in August 2019 to more cheaply hedge spot exposure and arbitrage price disparities with derivatives, the investor letter says. It will also initiate cryptocurrency options trades on the Deribit exchange in the fourth quarter of this year, but the plans are tentative to exercise caution over low and therefore risky options volume. Deribit alone trades less than $50 million a day.
Galois Capital has done some notable long-short trades, such as going long bitcoin at $3,750 in December 2018 and buying the FTX crypto-derivatives exchange’s FTT coin at $0.10 in April 2019, according to the letter.
The firm took the optimistic bitcoin position at a market low on the view that industry-wide hedge fund investment redemptions had subsided, that a crash in initial coin offerings — virtual currency investment structures highly correlated to bitcoin — had bottomed out and that legal measures surrounding Mt. Gox bankruptcy proceedings had abated bitcoin selling pressure.
And, due to a close relationship with FTX’s parent company Alameda Research, Galois Capital saw FTT as undervalued at the time of its investment. The coin sale was “very rushed and did not tap all of the available capital” despite “a surplus of demand on the sidelines,” the investor letter says. Galois Capital exited the FTT long position at a price between $0.80 and $1.94 while retaining some holdings of the coin.
In the coming year, Galois Capital will trade against its own market-maker, combining its liquidity and hedge fund services. Borrowed from traditional prop trading shops Two Sigma, Jump Trading and Tower Research, the strategy is aimed at improving long-short trading efficiency. According to the letter, Galois long-short traders have accidentally taken opposite positions and strive to trade independently without confusing each other’s profits and losses.