Renaissance Technologies is the Long Island firm whose Medallion fund produced the best long-run returns on record, then closed its doors to every outsider. This is the story of the house, not the man: for the founder himself, read our profile of Jim Simons. Episode two of The Great Houses.
James Simons wrote to the New York Times in 1967 to say that his boss did not speak for him, and within months it had cost him the best job in American mathematics.
He was twenty-nine, a Berkeley PhD who had finished his doctorate at twenty-three, and he worked at the Institute for Defense Analyses in Princeton, a Cold War outfit where elite mathematicians broke Soviet ciphers behind doors that officially did not exist. His classified work was considered brilliant by colleagues who could not describe it.
The problem was Vietnam. The president of IDA, Maxwell Taylor, had defended the war in print. Simons wrote to the paper to dissent, then told a magazine reporter he intended to work only on his own projects until the war was over.
In a secrets business, the offence is not the opinion. It is saying it in print.
They let him go. Simons walked out of the most sophisticated pattern-detection shop in the Western world, unemployed, with a wife, three children, and a conviction it would take him another decade to act on: the methods used to find signal inside enemy noise might find signal inside other kinds of noise as well.
The market, for instance, which is the noisiest system ever built by man, and which for the next half century will be systematically stripped for signal by the people this story is about.
From Codebreaker to Currency Trader
Simons was never the standard picture of anything. He grew up outside Boston, the son of a shoe factory manager, went to MIT at seventeen, and rode a motor scooter from Boston to Bogotá after graduating because the idea appealed to him. The mathematics, though, was entirely serious. His doctoral work on the geometry of curved spaces led, years later, to a collaboration with the geometer Shiing-Shen Chern that produced the Chern-Simons invariants, mathematics that theoretical physicists would spend decades applying to problems Simons never intended it for. In 1976 the American Mathematical Society gave him the Veblen Prize, the field’s top award in geometry. By then he had also rebuilt the mathematics department at Stony Brook on Long Island, recruited to run it at thirty, and turned a commuter school’s department into a serious research centre by hiring the way he would later hire traders: find the best person, pay what it takes, leave them alone.
He had also, all along, been gambling. Simons had put money into a flooring business in Colombia with friends and traded commodities on the side through the 1970s, and the itch outlasted every academic honour. In 1978, at forty, he left one of the most secure positions in American mathematics, took an office in a strip mall in Setauket, Long Island, next to a pizza restaurant and a women’s clothing shop, and set up a currency trading firm he called Monemetrics. Colleagues in the mathematics world treated it as a midlife aberration. Some told him so.
The plan, from the beginning, was models. The practice, for the first several years, was not, and the gap between the two is where the firm was actually made, because the man who closed it had to fail at ordinary trading first.
Simons recruited Leonard Baum, an IDA colleague and co-creator of the Baum-Welch algorithm, a method for teasing hidden states out of observable data that would later underpin speech recognition. Baum was one of the few people alive whose mathematics was genuinely suited to markets. And Baum, arriving, discovered he enjoyed something else entirely: trading on judgment. He bought currencies on instinct and macro reasoning, made money, and drifted further from the models with every winning month.
For a while Simons drifted with him. The firm rode the dollar and the peso and the yen on conviction, and Simons discovered that discretionary trading came with a cost the mathematics never had. Positions kept him awake. Losses reached him physically. He would later describe the experience as gut-wrenching, a man accustomed to proof discovering what it is to be long and wrong with no theorem to appeal to.
There had been a real model, early. Simons and Baum had built a system that tracked trending behaviour in currencies, and for a stretch it worked well enough that the pair could have simply kept refining it. Instead its own success became the argument against it: Baum, watching the model’s positions, concluded he could see what it saw and more, and that the machinery was an unnecessary intermediary between his judgment and the market. It is the oldest failure pattern in systematic trading, and it claimed one of the founders of the mathematics of hidden states. The model was set aside. The judgment took over.
The drift ended the way it usually ends. In the mid-1980s Baum’s instinct-driven bond positions went against him hard, the losses ran into millions, and Baum left the firm that his own algorithm should have defined. He returned to mathematics, spent his later decades on number theory, and remains the first name in this story’s ledger of casualties: the man who carried the right tools to the right door and chose not to use them.
Simons kept the conclusion. Judgment did not scale, did not sleep, and did not survive its own drawdowns. Whatever this firm became, no human being would be trusted with the positions again.
Four Mathematicians Build Medallion
The firm that emerged, renamed Renaissance Technologies in 1982, was built by a sequence of mathematicians, each of whom solved one layer of the problem and most of whom did not survive the solving.
The first layer was data, and the man who owned it was Sandor Straus. In the early 1980s, price history was an afterthought even at serious firms: patchy, error-ridden, bought on tapes and trusted on faith.
Straus treated it as the asset. He collected tick data nobody else wanted, cleaned it by hand, hunted down errors one misprinted price at a time, and built, year over year, what became one of the most complete records of market behaviour in existence. He gathered opening and closing prices going back decades, then intraday ticks when almost no one stored them, buying old records, transcribing paper archives, writing programs to catch the impossible prints, the decimal slips, the days that appeared twice. When colleagues questioned the expense of hoarding data at resolutions no strategy yet used, Straus kept hoarding, on the theory that the strategies would come and the history could not be bought later at any price. He was right about both.
It was invisible work of the kind no magazine profiles, and it was the foundation everything else stood on: a model is only as honest as the history it is trained against, and Renaissance’s history was cleaner than anyone’s, at horizons no rival could reconstruct. Straus is this story’s Engine, and like most Engines he is the least famous person in it.
The second layer was a working model, and it came from James Ax, a number theorist Simons had hired at Stony Brook and now hired again. Ax took the firm’s research and Straus’s data and built trading systems good enough that in 1988 Simons and Ax launched a dedicated fund around them. They named it Medallion, for the medals each had won. It would become the most profitable investment vehicle in the history of money, and in its second year it nearly died.
Through 1989 the models bled month after month, in a drawdown later reported at close to 30 per cent. The two men read the same numbers and reached opposite conclusions. Ax, working from his compound on the California coast, held that the losses sat within the system’s statistical expectations and that intervening in a model because it was losing was precisely the error the firm existed to avoid. Simons held that a theory of tolerable loss is itself a model, and that this one was failing the only test that mattered.
The argument turned bitter before it resolved the only way it could: Simons owned the larger share, trading was halted, and the systems went back to the workshop. Ax left the firm and, largely, the industry. The dispute was the founding constitutional crisis of Renaissance, and it settled the question every quant firm eventually faces: when the model and the owner disagree, someone must have the final key. Simons kept it.
The third layer was the fix, and it belonged to Elwyn Berlekamp, a Berkeley coding theorist and games expert who had bought into Ax’s operation and now took over its repair. Berlekamp pushed the system toward shorter horizons and higher frequency, toward thousands of small positions held briefly rather than convictions held long, on the logic that a small statistical edge, exercised constantly, compounds into certainty while a large edge exercised rarely remains a gamble. In 1990 the rebuilt Medallion returned over 55 per cent after fees. Berlekamp, having proven the point, sold his stake back and returned to Berkeley, one of the few figures in this series to walk away from a money machine at its ignition because he found the machine less interesting than the mathematics.
The fourth layer was doctrine, and it came from Henry Laufer, another Stony Brook mathematician, who pressed the firm toward a single unified model: not a drawer of separate strategies but one integrated system in which every signal, every market, every position size was computed jointly, so that the whole portfolio was one continuously re-solved problem. Rival firms ran collections of models and let them fight. Renaissance ran one brain.
By the early 1990s the brain was printing money in futures and currencies. What it could not yet do was trade the largest market on earth. Equities, thousands of names, dividends, splits, borrow costs, market impact, kept defeating the system, and the two men who finally solved it came, fittingly, from the same lineage Simons himself did: hidden signal inside noise.
Brown, Mercer and the Equities Problem
Peter Brown and Robert Mercer arrived together from IBM in 1993. At IBM they had been central figures in the statistical revolution in speech recognition and machine translation, the then-heretical idea that a computer should not be taught grammar but should instead infer language from brute statistical exposure to millions of examples. The linguists hated it. It worked. Translation, it turned out, was a decoding problem, which is to say the same problem Simons had worked at IDA and the same problem Baum’s algorithm addressed, and when Renaissance’s recruiter, Nick Patterson, himself a former code-breaker, came calling, the fit was closer to a homecoming than a career change.
Brown was kinetic, competitive, and slept famously little, keeping a fold-down bed in his office wall and working past the point where colleagues went home twice; Mercer was so silent that colleagues joked about rationing his words, a man who could sit through an hour’s meeting and spend nothing. They had been a unit at IBM for years, the talker and the silence, and Renaissance hired the unit.
Together they attacked the equities problem for two years while it attacked back. The system they built lost money in live trading for reasons no one could locate, and the firm came close to abandoning equities entirely before a young programmer named David Magerman, digging through the code, found the errors, one of them sitting in a line Mercer had written.
The fixed system worked.
Then it more than worked. Equities gave Medallion what futures never could: thousands of instruments across which to spread thousands of small, statistically tilted bets, the law of large numbers turned into an industrial process.
The philosophy underneath was stated inside the firm with unusual honesty. Renaissance did not believe it could predict the market. It believed it could be right slightly more often than wrong, something in the region of 50.75 per cent of the time by Mercer’s own reported account, and that at sufficient scale and frequency, that sliver is not a small edge but an absolute one. The signals themselves were frequently inexplicable. Some patterns made economic sense; many made none, and the firm’s discipline was to trade what tested true rather than what could be narrated. In a century of finance built on stories, Renaissance’s founding act was to stop requiring them.
Inside the Medallion Fund
The firm has never said what the machine actually traded. The honest outside answer assembles from fragments. Mean reversion at short horizons, the tendency of prices pushed too far too fast to breathe back. Patterns keyed to time of day and day of week, the mechanical rhythms of other people’s order flow. The predictable footprints of large investors moving size. Relationships between instruments that held often enough to bet on, and weakly enough that nobody else bothered. None of it resembled an investment thesis. All of it resembled what a signals-intelligence unit does to an enemy transmission, which is the biographical joke of the firm: Simons, Baum, Patterson, and the IBM men had all spent formative years extracting meaning from streams specifically engineered to hide it. The market, which hides its meaning by accident rather than design, never stood the same chance.
The numbers that resulted have been checked, disputed, and re-checked for thirty years, and the consensus reconstruction is this: from 1988 to 2018 Medallion returned roughly 66 per cent a year before fees and roughly 39 per cent after them, with fees that themselves became legend, 5 per cent of assets and, eventually, 44 per cent of profits. Total trading profits over the period exceeded $100 billion, by the reconstruction in Gregory Zuckerman’s history of the firm, The Man Who Solved the Market. It never had a losing year in that span. In 2008, with the financial system on fire around it, the fund returned about 82 per cent after fees. The joke inside the industry was that Medallion’s fee schedule was the most expensive in the world and underpriced.
The design choice that made it possible was the one almost no one on Wall Street has ever willingly made: staying small. The signals Medallion traded decayed with size; push too much money through them and the act of trading destroyed the pattern being traded.
The arithmetic is worth one paragraph, because it explains the fact outsiders find hardest to believe. A signal that predicts a price move of a few hundredths of a per cent can absorb only so much capital before the buying itself moves the price more than the predicted move, at which point the trade eats its own edge. Every strategy therefore has a capacity, and Medallion’s, given its horizons, was measured in single-digit billions.
Wall Street’s universal response to a working strategy is to scale it until it stops working and charge fees on the corpse. Renaissance’s response was to cap the fund at roughly $10 billion, force profits out annually rather than letting them compound inside, and, by 2005, buy out the last outside investors entirely. From then on Medallion belonged only to the people who built and ran it. The most reliable money machine in financial history closed its doors not out of exclusivity as marketing, but because exclusivity was a load-bearing wall.
Around the machine grew a campus. Renaissance moved to a wooded compound in East Setauket that resembled a mathematics institute more than a trading firm, some three hundred employees, a third with PhDs, drawn from physics, statistics, astronomy, and computational linguistics, and almost never from Wall Street. The hiring rule was absolute: the firm wanted people who had done first-rate original science and it did not want people who had worked in finance, on the stated logic that a Wall Street background meant habits to unlearn and, worse, opinions about markets. An astronomer who had cleaned signal from telescope noise, a linguist who had modelled the probability of one word following another, arrived with the right instincts and no folklore.
Internally the culture ran on open science: one shared codebase, results presented to the whole research floor, compensation tied to the single fund everyone ate from. Externally it ran on silence. Employees signed agreements that kept them from ever working in the industry again, and the firm published nothing, explained nothing, and confirmed nothing. The contrast was the point. Secrecy toward the world, glasnost within: the exact inversion of the typical bank, where traders guard their books from the desk beside them and the firm advertises itself to everyone else.
The secrecy was not a pose; it was enforced like a perimeter. In 2003 two researchers, Alexander Belopolsky and Pavel Volfbeyn, left for Millennium Management, the pod-shop empire, and Renaissance sued them and their new employer, alleging its trading secrets had walked out the door. The fight ran for years before it was settled. The message to three hundred employees who had signed lifetime restrictions was received without needing a memo: the mathematics belonged to the house.
The house had also, by then, survived the comparison that defined its era. In 1998 Long-Term Capital Management, the other famous fund built on mathematics, run by Nobel laureates and financed on the certainty of convergence, detonated so spectacularly that regulators feared for the banking system. For years afterward, quantitative meant dangerous in the mouths of allocators. Renaissance traded through the episode profitably and drew the distinction that mattered: LTCM had used mathematics to justify enormous positions in a small number of ideas it believed could not fail. Renaissance used mathematics to hold thousands of positions it fully expected to fail almost half the time. Humility about any single trade, arrogance about the aggregate: the inversion of Greenwich.
The machine’s worst week came in August 2007, when a forced liquidation somewhere in the quant world set off a stampede through the crowded statistical-arbitrage trades, a story this series tells in full in the AQR episode.
Inside Renaissance the episode produced the rarest of events: a human override. As the losses accelerated through that August week, with even Medallion bleeding at a pace its history said was close to impossible, Simons ordered positions cut. Brown and Mercer argued the other side, that the models had earned their trust precisely for moments like this one, and that selling into the stampede made the firm part of the stampede. Simons’s answer, in substance, was that survival is senior to statistics: the models assumed the firm would be there tomorrow, and his job was to make the assumption true.
The markets snapped back within days, the unwound positions recovered without Renaissance aboard, and the models were vindicated at the exact moment their master overrode them. Both men were right, which is why the episode is still argued about in quant circles, and the firm absorbed the harder lesson underneath: the resolve to trust the machine is itself maintained by fallible people with stomachs, and no amount of mathematics removes the one seat at which a human being decides whether the mathematics still applies.
Taxes and Politics
The 2010s brought the two controversies that mark the house, and both are matters of public record.
The first was about taxes. For years, Medallion had executed much of its rapid-fire trading through basket option structures arranged with Deutsche Bank and Barclays, instruments under which the banks technically held the positions and the fund held an option on the basket’s performance. The economic effect was Medallion’s trading; the tax effect was that profits generated in days could be reported at long-term capital gains rates, and leverage could run beyond what direct ownership would have allowed. The scale was industrial: over more than a decade the structures processed hundreds of billions of dollars of trading volume, and the subcommittee’s staff calculated the tax saved in the billions.
In 2014 the Senate’s Permanent Subcommittee on Investigations put the structure through a day of public hearing. Peter Brown sat in the witness chair alongside the banks that had manufactured the product. Senator Carl Levin, running his last great inquiry, pressed the obvious point. A fund that held positions for days, sometimes hours, was reporting its profits as the fruit of long-term patience. The subcommittee concluded the options were a device for buying a tax treatment the underlying trading could never have earned.
The firm defended their legality. The argument ran seven more years, and in September 2021 current and former Renaissance executives agreed to pay as much as $7 billion in back taxes, interest, and penalties to the IRS, reported at the time as one of the largest tax settlements in American history. Simons paid a further $670 million on top. Peter Brown told investors the firm had settled rather than risk a worse outcome in litigation. No fraud was charged.
The second was about politics, and it made the firm’s quietest man its most famous. Robert Mercer had, through the 2010s, become one of the largest financial forces in American right-wing politics, funding Breitbart News, backing Stephen Bannon, and, with his daughter Rebekah, becoming a decisive early patron of Donald Trump’s 2016 campaign, with the data firm Cambridge Analytica among the family’s investments. None of this was done in Renaissance’s name, and Simons himself was among the Democratic Party’s largest donors, a symmetry the firm pointed to when asked.
But the attention was corrosive in ways a secrecy-built institution was unequipped for. David Magerman, the programmer who had once fixed Mercer’s code and spent two decades at the firm, went public against Mercer’s politics in February 2017, criticising his co-chief executive’s support for Trump in the Wall Street Journal. He was suspended without pay. At a charity poker tournament in Manhattan that April, by his own later court complaint, Rebekah Mercer confronted him and called him pond scum. He was fired nine days later, sued Mercer for wrongful dismissal, and dropped the suit that summer. He became the rarest figure in the Renaissance story: an insider who talked.
Clients of the public funds raised the association. In November 2017 Mercer announced he would step down as co-chief executive, and Peter Brown ran the firm alone. The record does not tell us what was said between the principals, and this series does not invent it.
The Door Outsiders Could Not Open
There was one boundary the machine itself drew, and the firm was unusually honest about it. Medallion’s edge lived at short horizons and limited size, and by the mid-2000s the firm faced the question its own success created: what to do with a research engine, a data estate, and a brand worth billions, attached to a fund that could not grow. The answer, launched in 2005, was a second family of funds for outsiders, RIEF and its siblings, engineered for capacity rather than velocity: slower signals, longer holding periods, room for tens of billions. Renaissance said plainly, in documents and in meetings, that these were different animals and not Medallion, that their returns would be good rather than miraculous, and investors, hearing the disclaimer from the house of miracles, subscribed to the name and filed the disclaimer.
In 2020 the difference stopped being theoretical. Through the pandemic convulsions, the public funds lost heavily, RIEF finishing the year down roughly 20 per cent, while Medallion, trading the same chaos at its own horizon, returned some 76 per cent, according to figures reported by Institutional Investor. Billions in outside money left in the redemptions that followed. Nothing about the divergence contradicted anything the firm had said for fifteen years, which is precisely what made it so clarifying: the edge was real, the edge was bounded, and the boundary sat exactly at the door where outside money came in.
Simons had stepped back from the chief executive role in 2010, handing the firm to Brown and Mercer, and left the chairmanship in 2021. The final act ran on the other ledger. Through the Simons Foundation, built with his wife Marilyn, he directed billions into mathematics, physics, computational biology, and autism research, and through Math for America he paid stipends to hundreds of public school mathematics teachers in New York on the unfashionable theory that the way to honour a subject is to pay the people who teach it. The scale made him one of the largest private patrons of fundamental research on earth, and the method was recognisably the Stony Brook method and the Renaissance method applied a third time: find the best people, fund them properly, leave them alone. He had run the same strategy on mathematicians, traders, and scientists, and it had never once lost money.
The philanthropy also carried a private weight the profiles handled carefully. Simons buried two of his sons: Paul, killed on a bicycle near the family’s home in 1996 at thirty-four, and Nick, drowned off Bali in 2003 at twenty-four. The family’s response to the second loss was characteristic of the man’s entire method, grief converted into structure: Nick had worked in Nepal, so the Simonses built the Nick Simons Institute there, training rural health workers in his name. Simons spoke of the deaths rarely and briefly. The record shows the institutions; the rest he kept. He died in Manhattan on the 10th of May, 2024, at eighty-six. The obituaries led with the returns. The mathematicians led with the geometry.
At the firm in East Setauket, the week he died, the models priced, the positions turned over, and the fund traded as it had the week before. Every other house in this series was mortal in the same place: the founder’s judgment, sitting unhedged at the centre of the book. Simons spent forty years engineering that judgment out of the system, replacing himself so thoroughly that his own death registered nowhere in the machine he built.
It is the only immortality this industry has ever produced, and it belongs to the man who was fired for saying what he thought.
At a Glance
| Key Fact | Detail |
|---|---|
| Founded | 1978 as Monemetrics, Long Island; renamed Renaissance Technologies in 1982 |
| Founder | James Simons, mathematician and former IDA codebreaker |
| Headquarters | East Setauket, New York |
| Flagship fund | Medallion, launched 1988 |
| Medallion returns | About 66% a year before fees and 39% after, 1988–2018 (Zuckerman reconstruction) |
| Fees | 5% of assets and up to 44% of profits |
| Outside investors | Bought out of Medallion by 2005 |
| Funds for outsiders | RIEF and siblings, launched 2005; RIEF lost roughly 20% in 2020 while Medallion gained about 76% |
| Tax settlement | Up to $7 billion paid to the IRS by current and former executives, 2021 |
| Founder’s death | 10 May 2024, Manhattan, aged 86 |
Keep Reading
- Jim Simons: The Mathematician Who Broke the Market
- Edward Thorp: The First Quant
- Jaffray Woodriff: The Quant Who Found a Third Way to Trade
- The Greatest Traders: the book
The Great Houses continues every Wednesday: one firm a week, from the Medici Bank to Jane Street. For the traders who built and fought these firms, start with The Greatest Traders.
The Complete Trader's Edge
The full Mind · Method · Money framework. 70 chapters.
View on Amazon →
Market Mayhem
400 years of bubbles, crashes, and the pattern that keeps repeating.
Buy on Amazon →
Greatest Companies
How the world's greatest companies were built — and what traders learn from them.
View on Amazon →
Greatest Traders
Eighty-six lives that explain the markets — Livermore to Madoff, told with the losses left in.
View on Amazon →



