Illustrator: Martin Ansin/Bloomberg Markets Inside a Moneymaking Machine Like No Other The Medallion Fund, an employees-only offering for the quants at Renaissance Technologies, is the blackest box in all of finance. Katherine Burtonburtonkathy November 20, 2016 — 11:01 PM CST Sixty miles eastof Wall Street, a spit of land shaped like a whale’s tail separates Long Island Sound and Conscience Bay. The mansions here, with their long, gated driveways and million-dollar views, are part of a hamlet called Old Field. Locals have another name for these moneyed lanes: the Renaissance Riviera. That’s because the area’s wealthiest residents, scientists all, work for the quantitative hedge fund Renaissance Technologies, based in nearby East Setauket. They are the creators and overseers of the Medallion Fund—perhaps the world’s greatest moneymaking machine. Medallion is open only to Renaissance’s roughly 300 employees, about 90 of whom are Ph.D.s, as well as a select few individuals with deep-rooted connections to the firm. The fabled fund, known for its intense secrecy, has produced about $55 billion in profit over the last 28 years, according to data compiled by Bloomberg, making it about$10 billion more profitablethan funds run by billionaires Ray Dalio and George Soros. What’s more, it did so in a shorter time and with fewer assets under management. The fund almost never loses money. Its biggest drawdown in one five-year period was half a percent. Illustrator: Martin Ansin “Renaissance is the commercial version of the Manhattan Project,” says Andrew Lo, a finance professor at MIT’s Sloan School of Management and chairman of AlphaSimplex, a quant research firm. Lo credits Jim Simons, the 78-year-old mathematician who founded Renaissance in 1982, for bringing so many scientists together. “They are the pinnacle of quant investing. No one else is even close.” Few firms are the subject of so much fascination, rumor, or speculation. Everyone has heard of Renaissance; almost no one knows what goes on inside. (The company alsooperates three hedge funds, open to outside investors, that together oversee about $26 billion, although their performance is less spectacular than Medallion’s.) Apart from Simons, who retired in 2009 to focus on philanthropic causes, relatively little has been known about this small group of scientists—whose vast wealth is greater than the gross domestic product of many countries and increasingly influences U.S. politics1—until now. Renaissance’s owners and executives declined to comment for this story through the company’s spokesman, Jonathan Gasthalter. What follows is the product of extensive research and more than two dozen interviews with people who know them, have worked with them, or have competed against them. Renaissance is unique, even among hedge funds, for the genius—and eccentricities—of its people. Peter Brown, who co-heads the firm, usually sleeps on a Murphy bed in his office. His counterpart, Robert Mercer, rarely speaks; you’re more likely to catch him whistlingYankee Doodle Dandyin meetings than to hear his voice.2Screaming battles seem to help a pair of identical twins, both of them Ph.D. string theorists, produce some of their best work. Employees aren’t above turf wars, either: A power grab may have once lifted a Russian scientist into a larger role within the highly profitable equity business in a new guard vs. old guard struggle. For outsiders, the mystery of mysteries is how Medallion has managed to pump out annualized returns of almost 80 percent a year, before fees. “Even after all these years they’ve managed to fend off copycats,” says Philippe Bonnefoy, a former Medallion investor who later co-founded Eleuthera Capital, a Switzerland-based quantitative macro firm. Competitors have identified some likely reasons for the fund’s success, though. Renaissance’s computers are some of the world’s most powerful, for one. Its employees have more—and better—data. They’ve found more signals on which to base their predictions and have better models for allocating capital. They also pay close attention to the cost of trades and to how their own trading moves the markets. But as computing power becomes ever cheaper and competitors sharpen their skills, will Medallion continue to mint money? Quants seem like saviors to investors disappointed with how mere mortals have managed their money of late. In 2016 clients plugged $21 billion into quant hedge funds, while pulling $60 billion from those that do everything else. One noteworthy quant shop, Two Sigma, managed just $5 billion during the financial crisis and has seen assets jump to $37 billion. Even old-fashioned traders such as Paul Tudor Jones and Steve Cohen are adding to their computer scientist ranks in hopes of boosting returns. Renaissance’s success, of course, ultimately lies with the people who built, improved upon, and maintain Medallion’s models, many of whom met at IBM in the 1980s, where they used statistical analysis to tackle daunting linguistic challenges. This is their story. Simons is alreadywell-known: math genius, professor at MIT and Harvard, recipient of the Oswald Veblen Prize in Geometry, and co-creator of the Chern-Simons theory. He was also a code breaker for the Institute for Defense Analyses, where he worked finding messages amid the noise. The goal of quant trading is similar: to build models that find signals hidden in the noise of the markets. Often they’re just whispers, yet they’ll help predict how the price of a stock or a bond or a barrel of oil might move. The problem is complex. Price movements depend on fundamentals and flows and the sometimes irrational behavior of people who are doing the buying and selling. Although Simons lost the IDA job after denouncing the Vietnam War ina letterto theNew York Times, the connections he made through his work in cryptography helped create Renaissance and, a few years later, Medallion. Over the next decade, while chairing the math department at Stony Brook University, Simons dabbled in trading commodity futures. In 1977 he left academia for good to try his hand at managing money.3 Initially he bought and sold commodities, making his bets based on fundamentals such as supply and demand. He found the experience gut wrenching, so he turned to his network of cryptographers and mathematicians for help looking at patterns:Elwyn Berlekampand Leonard Baum, former colleagues from IDA, and Stony Brook professors Henry Laufer and James Ax. “Maybe there were some ways to predict prices statistically,” Simons said in a 2015 interview with Numberphile. “Gradually we built models.” At their core, such models usually fall into one of two camps, trend-following or mean-reversion. Renaissance’s system had a foot in both. Its results were mixed at first: up 8.8 percent in 1988, its first year, and down 4.1 percent in 1989. But in 1990, after focusing exclusively on shorter-term trading, Medallion chalked up a 56 percent return, net of fees. “I was confident that the models would work better,” says Berlekamp, who returned to academia in 1991 and is now a professor emeritus at the University of California at Berkeley. “I didn’t think they would be as good as they were.” Eventually the scientists went so far as to develop an in-house programming language for their models rather than settle for a numbercentric option such as ASCII, which was popular at the time. Today, Medallion uses dozens of “strategies” that run together as one system. The code powering the fund includes several million lines, according to people familiar with the company. Various teams are responsible for specific areas of research, but in practice everybody can work on everything. There’s a meeting every Tuesday to hash out ideas. In the early 1990s,big annual returns became the norm at Renaissance: 39.4 percent, 34 percent, 39.1 percent. Prospective investors clamored to get into Medallion, but the company didn’t pay them much heed—or coddle clients for that matter. Bonnefoy recalls dialing a Manhattan phone number to hear a recording of the monthly returns; Renaissance’s legal department doubled as unhelpful customer service representatives. (To this day the company’s website, rentec.com, looks like it dates from the Netscape era.) In 1993, Renaissance stopped accepting new money from outsiders. Fees were also ratcheted up—from 5 percent of assets and 20 percent of profits, to 5 percent and 44 percent. “They raised their fees to exorbitant levels and were still head and shoulders above everyone else,” says Bonnefoy, who, along with every other outsider, was finally booted from Medallion in 2005. Encouraged by Medallion’s success, Simons by the mid-’90s was looking for more researchers. A résumé with Wall Street experience or even a finance background was a firm pass. “We hire people who have done good science,” Simons once said. The next surge of talent—much of which remains the core of the company today—came from a team of mathematicians at the IBM Thomas J. Watson Research Center in Yorktown Heights, N.Y., who were wrestling with speech recognition and machine translation. In the early days of tackling these problems, computer scientists teamed with linguists and tried to code grammar. At IBM, a group including Mercer and Brown reasoned that the problems would be better solved using statistics and probabilities. (Their boss, Frederick Jelinek, liked to say, “Whenever I fire a linguist, the system gets better.”) According to scientists who worked at the research center then, the team fed reams of data into its computers. Documents from the Canadian Parliament, for instance, were available in both English and French, which none of the scientists spoke. (Mercer once disappeared for several months to type French verb conjugations into a computer, according to a source.) The data allowed them to write an algorithm that found the most likely match for the phraseLe chien est battu par Jean was “John does beat the dog.” A similar approach applied to speech recognition: Given auditory signalx, the speaker probably said the wordy. “Speech recognition and translation are the intersection of math and computer science,” says Ernie Chan, who worked at the research center in the mid-1990s and now runs quant firm QTS Capital Management. The scientists weren’t just working on academic problems; they were also developing theories and writing software to implement the solutions, he says. The group’s work eventually paved the way for Google Translate and Apple’s Siri. Mercer and Brown went to IBM’s management in 1993 with a bold proposition, says a person who knows the two: Let them build models to manage a portion of the colossal company’s then-$28 billion pension fund. IBM balked, questioning what computational linguists would know about overseeing investments. But the duo’s fascination with financial markets was just beginning. That same year,Nick Patterson, a former code breaker for British and U.S. intelligence agencies, joined Renaissance and approached acquaintances Brown and Mercer. “IBM was in serious trouble, and morale was poor, so it was something of a recruiting opportunity,” says Patterson, who worked at Renaissance until 2001 and is now a senior computational biologist researching genetics at the Broad Institute of MIT and Harvard. The two decided to join, drawn by the 50 percent pay raise. They roomed in an attic apartment in Setauket and often dined together. When the bill came, they would pull out a special calculator that could generate random numbers. Whoever produced the higher number picked up the tab. “Renaissance was started by a couple of mathematicians,” Brown said in a 2013 conference for computational linguists. “They had no idea how to program. They’re people who learned how to program by reading computer manuals, and that’s not a particularly good way of learning.” He and Mercer had learned how to build large systems—with many people working on them simultaneously—which was a skill set they used to Renaissance’s advantage. Not that their new field was without challenges. “It’s all noise in finance,” he said. More IBM veterans joined them on Long Island, including Stephen and Vincent Della Pietra, the string-theorist twins;Lalit Bahl, who had created algorithms to recognize human speech; Mukund Padmanabhan, whose specialty was digital-signal processing;David Magerman, a programmer; and Glen Whitney, who wrote software as a summer intern. “The takeaway from IBM was that the whole is greater than the sum of its parts,” says Chan. “They all worked together.” Renaissance also spent heavily collecting, sorting, and cleaning data, as well as making it accessible to its researchers. “If you have an idea, you want to test it quickly. And if you have to get the data in shape, it slows down the process tremendously,” says Patterson.... http://www.bloomberg.com/news/artic...-medallion-fund-became-finance-s-blackest-box
A client of mine was one of the first investor in Medaillon. However, neither him or anyone I know has ever seen an audited statement of the fund for Medaillon. Their fund started to perform badly after they raised a lot of money. They launched a 400m$ equity fund a few years back and they weren't able to make more than 4-5% of return per year, while allegedly making 50% p.a. on multiple billions. Bad mouths would say it was a ponzi scheme that ended well.
This seems to be a recurring theme in such ventures. Finding trading systems with an edge is hard enough, but finding profitable systems that scale up is nearly impossible. I have a couple of systems that work for a sub-million account but if I pushed it above that I'd start being a significant fraction of the open interest. "Here be dragons."
Could this be because the more money you apply to the market, the bigger effect and the bigger footprint you're leaving behind. This doesn't go unnoticed by algo's, particularly if you're repeating a setup over and over with large size. I imagine in order to make huge consistent returns, you'd have to have a way to track big money to at least some degree and after they make an action, you react on the footprint or clue they leave behind. A big part of intra-day trading is just people locating imbalances, applying their capital with a calculated risk to force others out. Like buying up supply and forcing shorts to cover, as the shorts cover and new buyers come in, the same people that bought are now selling to those being forced out. Don't know just a thought.
Katherine Burton is a good reporter but this is mostly a patch-together of previously reported material. RenT evidently didn't cooperate, nor did its ex-employees.
But surely scalability is an issue that PhDs can understand, manage and communicate to investors, as it is very often done in the hedge fund industry, strategies with limited capacity have a hard close at the given capacity. Additionally, the fund was doing 80% p.a. + back when it had 1Bn+ in asset so this doesn't explain why newly funded strategies with only 400m$ can't perform on such small amounts. Just playing devil's advocate, not saying it's a ponzi scheme.
It seems likely that the "public" fund does not use or have access to the entire scope of what is being used for the "in house" fund. Which would make some sense considering that public money had been kicked out of the in house fund completely some years ago. ..... Or it is just a ponzi scheme used to feed capital into the in house fund to maintain its "profitability".
http://techinpink.com/2016/09/30/ho...rket-at-numerai-ml-tournament-final-tutorial/ https://numer.ai/ machine learning.... each bar on any time series has 20 other dimensions that are predictive of the next bar. Its not just the high low close of each bar. Super computers are fed probably more than 20 parameters each second to predict the next bar. some parameters have a higher predictive value. How many parameters do you think a trader's brain takes into account to make a decision in the market, besides just the bars/bar timeframe.