How the Medallion Fund actually makes money

Start with the thing almost everyone gets wrong

Renaissance runs no mutual fund. It runs no ETF. There is no Renaissance product you or I can buy, and there never has been.

Its Form ADV, the disclosure every US investment adviser files, has a line for registered investment companies, which is the legal category holding every mutual fund and ETF in the country. Renaissance reports zero clients and zero dollars on that line, and the box for managing registered funds is unticked. Its filing history contains no fund registration statement at all, only quarterly holdings reports and private placement notices.

The name causes most of the confusion. Renaissance Capital, a Connecticut research firm, sponsors an IPO ETF that anyone can buy and that has nothing to do with Renaissance Technologies. Renaissance Investments is a Canadian fund family offered by CIBC. Renaissance Investment Management is a separate Kentucky adviser. Medallion Financial Corp is a New York lender that once financed taxi medallions and shares nothing with the Medallion Fund but the word.

This matters more than a naming mix‑up, because the reason there is no public product is the same reason the returns are what they are. The strategy has a hard capacity limit. Selling it to the public would destroy it, and the firm has effectively said so.

What the firm actually is

Jim Simons, a geometer who chaired the mathematics department at Stony Brook and before that broke codes for the Institute for Defense Analyses, started a currency trading shop in 1978. It became Renaissance Technologies in 1982, and the Medallion Fund began trading in 1988.

The research operation sits on a wooded campus at 600 Route 25A in East Setauket, twenty minutes from me on Long Island, where 154 of its 170 investment staff work. Its July 2026 filing reports 310 employees excluding clerical staff. The company website, whose figures carry no date and have not moved since at least 2014, claims 90 PhDs, more than 50,000 computer cores, and a research database growing by more than 40 terabytes a day.

There are three fund families plus an employee fund of funds, and the differences between them are the entire story.

Figures as reported at the 31 December 2025 annual update. Gross asset value is stated without deducting borrowings, so it reflects assets financed by leverage and is not investor capital. Minimums are stated policy, and the brochure allows lesser amounts at the firm's discretion.

Look at the ownership column. Medallion is about 80 percent owned by the adviser and its related persons. RIEF is 8 percent. Medallion closed to new outside money in 1993, and the last unaffiliated investors were bought out in 2005.

So the fee schedule is not what it appears. A 4 percent management fee plus up to 44 percent of profits would be predatory if outsiders paid it. Renaissance's own brochure explains what is really going on: those fees "are largely responsible for the amount of compensation that Renaissance employees receive." The staff pay the fees to the firm, and the firm pays them back as compensation. Medallion is not a product. It is a payroll.

The record, and how much of it is verifiable

From 1988 to 2018, Medallion's annual returns averaged 66.1 percent before fees and about 39 percent after. There was no losing year before fees in 31 years, and exactly one after fees.

That sentence needs a permanent caveat, so here it is first. Renaissance publishes nothing. Those numbers come from the appendix of Gregory Zuckerman's 2019 book, reconstructed by a journalist from sources in and around the firm. The finance professor Bradford Cornell reproduced the return columns in the Journal of Portfolio Management and did the arithmetic. No audited statement has ever been public. Everything below inherits that limitation.

Compounding those annual figures gives 63.3 percent a year before fees. One hundred dollars placed in Medallion at the start of 1988 becomes $398,723,873 by the end of 2018, against $1,910 for the market. Cornell, who has read performance papers for forty years, wrote that he had "never seen any performance approaching that reported by Medallion" and called it a Michelson‑Morley level challenge to the idea that markets are efficient.

One detail in Cornell's paper is easy to skim past and matters more than the headline. Those gross returns are already net of trading costs. As he puts it, that "makes Medallion's performance even more extraordinary. It also implies that Renaissance was apparently particularly effective in minimizing such costs." A fund turning over its book every few weeks pays an enormous amount to trade, and the 66 percent is what survived that bill. Execution is not a supporting detail here. It is part of the edge.

The fee band on the chart is worth a moment. Applying the appendix's own schedule, roughly $104.5 billion of gross trading profit through 2018 splits into about $56 billion credited to investor capital and about $49 billion taken as fees. Since the investors and the firm are the same people, that split is mostly an internal accounting question, which is why the gross line is the one that tells you about the machine.

After 2018 the record goes dark

Medallion's disclosure did not merely stop being audited, it stopped existing. Institutional Investor reported the fund up 76 percent in 2020 and 48 percent in 2021, both sourced to investors rather than the firm. Zuckerman posted that it returned about 30 percent in 2024 on roughly $12 billion. For 2019, 2022, 2023 and 2025 there is no credible public figure at all. Numbers circulate for those years, and tracing them leads to blogs, forum comments and aggregator pages, one of which has a chatbot session pasted into its source. Treat any confident quote of Medallion's 2025 return as invented.

The funds outsiders could buy tell a different story

This is the most useful comparison available, because it is the only apples‑to‑apples test that exists: the same firm, the same scientists, the same research process, applied to outside money. The Renaissance Institutional Equities Fund launched in July 2005 and has returned roughly 9.5 to 10 percent a year since, depending on the end date. Over every window that can be checked it has trailed the S&P 500 with dividends reinvested, with drawdowns of about 35 percent from 2007 to 2009 and 27 percent from 2020 to 2021.

In 2020, the year Medallion gained 76 percent, that fund lost about 19 percent and its sibling lost about 32 percent. Investors pulled about $14.6 billion over the following fourteen months, and external assets fell from $65.1 billion to $23.2 billion over five years. In October 2025 the equities fund fell 14.39 percent in a single month, its worst in more than a decade.

How the money is actually made

In 2014 the Senate's Permanent Subcommittee on Investigations spent a year examining Renaissance's use of bank options, reviewed 1.5 million pages and put its executives under oath. That hearing is the richest public description of the machine that exists, because the firm had to explain on the record why it needed twenty to one leverage. Five things are doing the work, and none of them is a secret formula.

1. The edge per trade is almost nothing. Renaissance told the Senate in writing that its model "makes predictions that are profitable only slightly more often than not," and that those predictions "can be easily overwhelmed by external events." The famous number attached to this is Robert Mercer's, and its status is worth being precise about: Zuckerman reports it as something Mercer told a friend, not as anything the firm has said publicly. In that telling Medallion was right about 50.75 percent of the time. A coin landing your way 50.75 times in a hundred is not an investment thesis. It is a casino's house edge.
2. Volume turns that edge into near‑certainty. The firm's own explanation names the mechanism: the model "generates a large number of recommendations, so that by virtue of the mathematical principle known as the law of large numbers, the variability of the returns produced by the model is greatly reduced." Renaissance estimated 100,000 to 150,000 trades a day at each of two banks. An SEC examination found approximately 129 million trade orders placed in the year to March 2009, across nearly 10,000 instruments in 18 countries.
3. Positions are held for days, not years. A holding‑period chart prepared by Renaissance and found in the SEC's examination file shows that between January 2007 and March 2009, 86.97 percent of Medallion's tax lots closed within three months and 0.16 percent were held more than a year. The Senate noted positions lasting minutes, with the portfolio changing on a second‑to‑second basis.
4. Leverage multiplies a small return into a large one. This is the part most retellings skip. Because the edge is thin, Renaissance told the Senate that "the rate of return obtained by applying the recommendations to an unleveraged portfolio would be very small." A US margin account is capped at two to one by Regulation T. Renaissance bought basket options from Deutsche Bank and Barclays carrying ceilings of 18 and 20 to one, and by its own account to the IRS ran average leverage of 11.7 and 12.9 to one in 2005 and 2006. The options also capped the downside at the premium paid. In the firm's words: "No other investment structure of which we are aware provides both high leverage and loss protection."
5. The tax treatment was worth billions. Sixty of those options were held longer than a year and exercised between 2000 and 2013, paying out $45.3 billion against $11.0 billion of premiums. That $34.2 billion of profit, from trades mostly lasting weeks, was reported as long‑term capital gain. Subcommittee staff estimated the difference against short‑term rates at $6.83 billion. In September 2021 Renaissance told investors that current and former executives and other Medallion investors would pay as much as about $7 billion to settle with the IRS.

What goes into the model

Renaissance's Senate statement lists the inputs: "news stories, analysts' reports, energy reports, crop reports, weather reports, regulatory filings, accounting data, and, of course, quotes and trades from markets around the world." At TED in 2015 Simons said everything is grist for the mill except hem lengths. None of it is secret data. It is public data, collected earlier, cleaned better and kept longer than anyone else bothered to. The obsession is the oldest part of the firm: in the 1980s Sandor Straus bought historical tick data on magnetic tape and cleaned it by hand, when almost nobody used intraday data at all.

One model, not a library of strategies

Henry Laufer's decision in the 1990s was that Medallion would run a single unified model across every instrument rather than a collection of separate ones. Asked about it in 2022, Simons said: "We have only one model, one system, which is improved continuously on the basis of new signals and market behavior in general." He was answering about Medallion specifically, thirteen years after retiring, and the firm's own 2026 brochure speaks throughout of plural models. Peter Brown told the Senate the system had over a million lines of code and a thousand man‑years of work in it, and that changes ran at about one or two a week.

The signals themselves are often not interpretable, and that is deliberate. Brown, as quoted by Zuckerman: "If there were signals that made a lot of sense that were very strong, they would have long‑ago been traded out." Mercer said the signals they had traded for fifteen years "make no sense. Otherwise someone else would have found them." Simons was blunt about the method: "we do a lot of data mining. We have thousands of computers scouring data for information related to the movement of tradeable instruments. Sometimes what we find works, and sometimes it does not."

Trend‑following would have been great in the '60s, and it was sort of okay in the '70s. By the '80s, it wasn't. Jim Simons, TED 2015, on the strategy the firm started with and abandoned

That is the answer to anyone hunting for "the Renaissance strategy." There isn't one. There is a factory that keeps producing small, short‑lived statistical edges to replace the ones the market has competed away.

Execution is a first‑class problem

Barclays gave Renaissance direct market access through its own software with sub‑millisecond internal latency, and trades were placed and executed by computer with no human intervention on either side. Two details cut against the folklore. The brochure says the firm "is not obligated to solicit competitive bids or seek the lowest available commission costs" and may pay more than the cheapest broker for execution quality and securities lending. And it warns that Medallion's high turnover "can be expected to result in high transaction costs." They do not trade cheaply. They trade well, and they model the cost of trading as part of the signal.

Why they cannot scale it, and what that tells you

Medallion's capital has sat in the region of $9 to $12 billion for over a decade while producing six to nine billion dollars of gross profit a year. Bloomberg reported in 2016 that the firm caps the fund and distributes profits every six months instead of reinvesting. Asked at TED why the outside investors were removed, Simons gave a one‑clause answer: "we bought out all the investors because there's a capacity to the fund."

Capacity is not modesty, it is arithmetic. The strategy makes money from small, short‑lived price discrepancies whose size is fixed by the market, not by how much money you point at them. Double the capital and you either take the same profit on twice the base, halving the return, or you trade bigger and your own orders move the prices you were trying to exploit. Cornell drew the obvious conclusion from the outside funds' ordinary results: there is a scale limit on whatever generates Medallion's returns.

The academic literature agrees about this family of strategies. AQR's researchers, using nearly a trillion dollars of live trades, found that value, size and momentum survive trading costs at large size but that short‑term reversal strategies, the closest public analogue to what Medallion does, do not survive at reasonable size. Khandani and Lo documented the decay directly: a standard daily contrarian strategy returned 1.38 percent a day in 1995 and 0.13 percent by 2007.

So the shape of the industry follows. The strongest strategies are the small ones, and they end up capacity‑constrained and internal. D.E. Shaw's two flagship funds, up 18.5 and 28.2 percent in 2025, are both closed to new capital. Citadel handed back about $5 billion of 2025 profits rather than growing. The most profitable trading firms of all, Jane Street, Hudson River Trading, XTX, Jump and Citadel Securities, manage no outside client money whatsoever. What gets sold to the public is what has room: slower, capacity‑rich, lower‑return strategies.

What this leaves for someone with a normal account

I went looking for the part a person could copy. The honest answer has three parts: the core is not available at any price, the documented odds for individuals trying anyway are worse than most people expect, and there is a narrow band of things that genuinely do transfer.

That last line is the one that ends most attempts. David Bailey and Marcos Lopez de Prado, writing in the Notices of the American Mathematical Society, showed that testing enough variations of a strategy will produce a spectacular backtest by chance alone. Their number: with only five years of data a researcher can be seriously misled after trying as few as 45 configurations. Anyone with a laptop and a weekend will try more than 45. That is precisely why Renaissance's process needs a hundred scientists testing against decades of data and each other's scrutiny.

What actually happens to individuals who try

The last row is the one to sit with. Those are full‑time professionals with research staff, institutional data and execution, and nine in ten still lose to an index fund over a decade.

What genuinely transfers

Three things from the Renaissance record survive the drop from ten billion dollars to a normal account, and none of them is a trading strategy.

• The method, as discipline. Decide in advance, test honestly, count how many variations you tried, model your costs before believing a result, and prefer many small independent bets to one large conviction. That is transferable and valuable, and it is mostly a defense against your own conclusions.
• The cost obsession. Renaissance discards real signals that cannot clear transaction costs. At retail scale, costs, spreads, borrow fees, margin interest and short‑term tax are larger relative to any edge you find, not smaller. A strategy that looks good before costs and marginal after them is not a strategy.
• The recruitment insight. The people earning Medallion returns are employees, not investors. Renaissance hires mathematicians, physicists and computer scientists with no finance background at all and teaches them markets internally, because Peter Brown says that is easier than teaching mathematics to people who know markets. The documented path inside a machine like this is a research career, not a brokerage account.

None of that means systematic investing is closed to you. It means the part that makes Medallion extraordinary is the part that does not travel, and the part that travels is available in a low‑cost index fund anyone can buy. That is only an unsatisfying answer if the premise was that a technique was being withheld. The record says the technique was never the moat.

What nobody outside the firm knows

A piece like this is only as good as its willingness to mark its own edges.

• The returns are a reconstruction. Every Medallion figure from 1988 to 2018 traces to one appendix in one book. It is corroborated at points by contemporaneous reporting and the Senate record, and no credible source disputes it, but it is not audited data.
• Four recent years are blank. There is no reliable public figure for 2019, 2022, 2023 or 2025.
• The Mercer quote is second‑hand. The 50.75 percent line is attributed to Zuckerman's book and repeated everywhere. I could not trace it to a primary recording or transcript.
• The 2021 settlement has no public document. It was conveyed in a letter to investors, reported by the Wall Street Journal and reviewed by Reuters. No final total was ever published, and $7 billion is a reported ceiling.
• The trade counts and leverage figures are the firm's own numbers, supplied to the Subcommittee and the IRS and relayed in the Senate report.
• The model itself is entirely unknown. No signal, no weight, no line of the million‑plus lines of code has ever been public. Anything claiming to describe "the Medallion algorithm" is invention.

You do not have to take my word for the structural claims. The Form ADV is public and free at reports.adviserinfo.sec.gov. Item 5.D is where you will find zero investment companies against 14 pooled‑vehicle clients, Item 1.F gives the offices, and Schedule D lists the funds, their gross assets and their minimums. The 2014 Senate report is on hsgac.senate.gov and is the source for every trade count and leverage figure quoted here.

Medallion year by year, 1988 to 2018

Gross and net returns and fund size from Zuckerman's appendix as reproduced in Cornell (2020). The net figures there are a mechanical calculation from the gross return and the fee schedule, not reported net asset values. S&P 500 total return from the Berkshire Hathaway 2025 annual report.

Who is telling you this

I'm James Folk. I build streaming and mobile software for a living, most recently player and app work on Xfinity Stream, Peacock, discovery+ and Gospel Stream; the full record is at Roku channel development. I wrote this because Renaissance is usually filed under finance when it reads far more like a systems story: one model, a million lines of code, and an edge that only exists because the execution and the data pipeline are better than anyone else's. If you liked the way this is put together, the other piece here is what a Roku channel actually costs, which applies the same "show the real drivers" approach to a question people usually answer with a shrug.

This is research, not financial advice. I am not a licensed adviser, nothing here is a recommendation to buy or sell anything, and every fund figure is as reported by its source on the date given.