The Man Who Solved the Market – Jim Simons of Renaissance Capital
Books · Investing
(Read this in December 2019, adding to Notion in Feb 2021)
URL:
Highlights
Used data to find undervalued opportunities
Mathematicians and scientists understand markets far better than MBAs and seasoned business people
Simons had never taken a finance class and only dabbled in training until he turned 40
In 1980s, Simons results were so awful that he was forced to halt his investing. He also hired two mathematicians to work with him, but ended up getting in fights with them
Simons didn’t have a clue how to estimate cash flows, identify new products, or forecast interest rates. Instead, he focused on cleaning reams of pricing data
CHAPTER 1 (youth)
Started with investing in stocks. Then got bored and moved to commodity futures (soybeans). Then stopped after he got busy with his thesis and marriage
After college, started a company in Colombia in which he had stock, but didn’t spend much time on
CHAPTER 2 (life in academia)
Simons helped cracked Soviet code and doubled his professor salary while doing so. Along side this, He decided to start a company to electronically trade and research stocks. He also got his boss and one of his coworkers to join the company, named istar. But in the end; he failed to raise enough money and didn’t get the business off the ground
But he still searched for new ways to make money and spent his time examining the stock market. He developed a new stock trading system. He ignored the information most investors focused on (earnings, dividends, and corporate news). Instead, he focused on a small number of macroscopic variables.
Posited that the market had 8 possible states, including high variance (when stocks experience larger than average moves) and good (when stocks generally rise). These states were compiled by best fitting the available pricing data, not by economic theory
Simons was then fired for telling Newsweek that he won’t do any government work until the Vietnam War ended
Divorced after both he and his wife cheated
Colombian company was sold and Simons had some money. He put his money and convinced his partners to put money in an econometric model that was developed by his friend — one that used econometrics to forecast commodity prices. For instance, by observing crop production etc
In late 1970s, it seemed to Simon that a new and volatile era had begun. In 1978, he left Academia to start his own investment firm focusing on currency trading
CHAPTER 3 (early days of business)
Some believed that all possible information was already baked into market prices, and that only news (impossible to predict) could push prices higher or lower
Others believed that price shifts reflected investors efforts to react to new economic and corporate news
Simons thought that there was a lot of hidden data that wasn’t baked into market prices. Financial prices were like weather — somewhat random, but with larger climate dynamics underneath them. He wanted to build financial models to identify structure in financial markets
Wanted to identify signal from the noise of data
Simons was able to recruit Lenny Baum and James Ax to work with him
Baum thought that Markov Chains could be a reasonably good model for this. Hidden Markov Models can be used to get an idea of a systems implied rules from the data, even if the explicit rules stay hidden. Baum’s algorithm
At first, the firm made a lot of money on the pound and the yen. But lost money soon afterwards. They used a mixture of signals from algorithms and human judgement
Then, he tried to make a 100% automated system, looking at data up to 50 years prior to do so. Which metrics are leading indicators for other metrics? Do natural phenomena affect prices?
This didn’t quite work. Then, moved to a more traditional value investing model. Relied on processing news faster than others — translating an obscure French journal etc. This worked really well. He changed the companies name to Renaissance Technologies. Baum’s approach made $43M in profits, double their original investment
Baum had a lot of patience and a proclivity to hold on for a long time (not selling high). When there was a disagreement on whether or not to sell gold (which would’ve been the right decision), Baum was reluctant though eventually coerced
A similar thing happened with Coffee futures. Simons sold at the right time, but Baum didn’t let go
When Baum lost more than 40% of his holdings, an automatic clause was triggered and he left the company. He was fairly happy afterwards, and passed away at the age of 86 in 2017
CHAPTER 4 - 6 (late 1980s to early 1990)
James Ax (insanely smart but with a bad temper) — was a prominent partner at this time. Was one of the early developers of Number Theory
A professor (Carmona) developed a “black-box” algorithm for Renaissance (v similar to modern machine learning approach) that made buy/sell decisions for stocks and commodities
Another professor (Berlekamp) suggested improvements to the model. For instance, by using the probability of predictions as an indicator for how much stock to buy/sell
Ax started to lose a lot of money and also started relying more on his instincts instead of the models. Simons focused on dealing with investors
As Ax’s performance continued to decline, Simons stepped in to focus on trading. Ax and Simons had a falling out. Berlekamp offered to buy out most of Ax’s stake at this point
Berlekamp moved to a more short-term oriented strategy, making decision making totally automated.
He thought frequent, short-term trading would reduce risk. Buying and selling infrequently magnifies the consequences of each move. Mess up a couple of times, and your portfolio could be doomed. Make a lot of trades, and each individual move is less important (reducing overall risk)
With a slight statistical edge, the law of large numbers would be on your side as you trade.
Berlekamp also looked at how prices moved around day of week, as well as right before/after economic data was released
Simons wanted to get some of his instincts (like on oil or gold) incorporated in the firm’s buy/sell decisions, but Berlekamp held firm that decided that only the system will make decisions
Simons wanted to push the envelope further. He made more requests to expand the team, but more satellite dishes, and get better computers. He asked Berlekamp to chip in to pay for the new expenses. He thought that the company could do 80% returns in a year, while Berlekamp thought the max they could do was 30%
Berlekamp felt under pressure. He’d remained a part-time professor at Berkeley. He asked Simons to buy him out (since he valued the company far more than Berlekamp did). In Dec 1990, Simons purchased Berlekamp’s interest for cash. Other employees swapped their AxCom stakes for shares in Renaissance, which began to manage the Medallion fund
CHAPTER 7 - 8 (early 1990s)
There was a lot of competition emerging at this time. Plenty of people were trying to get in the market for automated trading as the quant movement emerged. Simons decided he needed help in order to stay competitive (specially as David Shaw and his firm were kicking ass)
Simons tried to get backing from other firms, but was unable to get it as he only had two years of good returns. By the time he had a third year of insanely good returns, some people thought he was running some kind of scam. The reason? He couldn’t explain why his models were making the decisions they were. Current investors were happy with their results, but very worried about the future
Financial economics worked, while it worked, largely because big investors felt compelled by academic theory to act “rationally”.
Simons and his colleagues believed that investors are prone to cognitive biases, while machines are not. Loss aversion, anchoring, and the endowment effect were all real.
They wanted to be super secretive. They even “misled” their investors and prospects into reasons for success, claiming that they used Perl (instead of C++, which others used) when they only used it for book-keeping
The mainstream didn’t trust Simons team, and in algorithms in general. They thought they were too risky
Some employees were happy with the firms AUM ($600M at this point), but Simons wanted more. The only conceivable way he thought they could to this was to go beyond commodities, and delve into equities
CHAPTER 9 - 10
There were a lot of bond and equity people who were making a killing at this time. The idea that computers could outperform they was thought of as ludicrous
His firm steered clear of Wall Street types. “We can teach your about money, we’re more interested in your approach”. I liked smart people who were probably unhappy in their current jobs, and didn’t have a familiarity with the industry
Patterson reached out to Peter Brown and Robert Mercer of IBM’s speech recognition group when IBM was cutting costs. They were also among the first pioneers in machine translation
Mercer had worked in a gov agency before, and thought that gov was useless. As was super focused and did mostly the same thing (clothes to wear, foods to eat) Everyday
Mercer and Brown cane over to RenTech in 1993, despite initial misgivings. Simons offered to 2x their salary. They worked on equities but weren’t able to make them profitable. Simons eventually gave them a year or so to try, but when it looked like it wasn’t working out, gave them a 6 month ultimatum
Mercer was calm, but Brown became a little neurotic. A junior programmer (megerman) then found some bugs in code being used to make decisions. He fixed them, the the program began making money soon afterwards
CHAPTER 11
The algorithms were consistently profitable, but still vulnerable to things like data entry errors. Also, the returns were slightly lower compared to previous years
Simons insisted on radical transparency. All code was visible to everyone else. There were no sections of the code only visible to top executives. We wanted then to swap ideas and collaborate with each other. Bonus depended on how well the firm performed, not by now your boss liked you. However, The firm became maniacal about not leaking outside the company, though
In the meanwhile, Quant investing took a reputational blow as another major firm (LTCM) went under and lost a lot of money
Medallion also started losing money because of a faulty strategy — by buying shares that had rallied in the previous week even when a vicious bear market had begun (and rallies were likely to be short lived). They soon fixed this
CHAPTER 12
From 2001 onwards, the Medallion fund became a data sponge. It quantified everything it could, and also started to work with more NLP and news. “There’s no data like more data”
Medallion started using Basket options (options linked to a basket of stocks)
Basket options also allowed Medallion to be eligible for lower Long Term Capital Gains Tax. Even though they made very frequent trades, they were eligible for lower taxes since the options were only exercised after a year
Brown and Mercer were promoted. Simons told them he wanted to he wanted to retire soon, and saw them as future co-CEOs
Simons increased management fees to 36% in 2002, and to 44% a bit later
In 2003, he began kicking investors out of Medallion and only kept employees. He didn’t want to make the fund too big
2 soviet-born employees began thinking of quitting in 2003. HR had fucked up when they became Principles. They had given them non-disclosure and non-compete agreements. The two hadn’t signed non-compete agreements, and HR hadn’t noticed
CHAPTER 13- 16
Simons retired and went into philanthropy. Brown and Mercer continued to lead the firm. New funds – RIEF and some others – also came into being and were successes