Jim Simons: the best record in history was never for sale
Every list of the greatest traders now puts Jim Simons at the top, and the number quoted is always the same: about 39% a year, after fees, for three decades. The number is real. What almost nobody adds is that you could never have bought it — and that the Renaissance fund you could buy spent fifteen years losing to an index fund.
KEY TAKEAWAYS
- Medallion: about 66% a year gross, 39% net, 1988–2021. Fees took roughly 41% of everything it earned — our division of the two reported figures.
- The edge was tiny. A Renaissance co-CEO put the hit rate at 50.75%. What was extraordinary was the repetition and the near-zero cost per trade.
- In 2000 Simons called $70bn inconceivable for his firm. By January 2020 the three funds open to outsiders held about $65.3bn — and then had their worst year ever.
- $10,000 in RIEF from launch to end-2020 became about $33,200. The S&P 500 turned the same money into about $41,400.
Who was Jim Simons?
A mathematician first, and for a long time nothing else. Simons took a PhD at Berkeley at 23, taught at MIT and Harvard, broke codes for the Institute for Defense Analyses until he was fired in 1968 for criticising the Vietnam war in print, then chaired the maths department at Stony Brook for eight years. His 1974 work with Shiing-Shen Chern — Chern–Simons theory — won geometry’s highest honour and is still used daily in theoretical physics.
He launched Medallion in March 1988, from a low wood-and-glass building on Long Island staffed with astrophysicists, number theorists and speech-recognition researchers. He was blunt about who he did not want: “We don’t hire people from business schools. We don’t hire people from Wall Street.” He died on 10 May 2024, aged 86. What earns him a page on a site about risk is not the brilliance. It is that he said out loud, in 2000, exactly where his method would stop working — and then watched his own firm walk past that line.
What did Medallion actually return?
Two numbers are quoted everywhere, both tracing to Gregory Zuckerman’s reporting in The Man Who Solved the Market (2019) and since repeated by Institutional Investor and the Wall Street Journal: from 1988 to 2021, Medallion returned about 66% a year before fees and 39% after. Individual years that leaked out include +98.5% in 2000, +82.4% in 2008 and +76% in 2020.
Start with the gap, because it is the part nobody frames properly. Sixty-six down to thirty-nine means fees consumed 27 percentage points, about 41% of everything the machine produced. Renaissance charged 5% of assets and, from 2002, 44% of profits. Even the best engine ever built was throttled by its own cost structure before a client saw a cent.
How big was the edge?
Small. Almost embarrassingly small.
We’re right 50.75 percent of the time… but we’re 100 percent right 50.75 percent of the time. You can make billions that way.Robert Mercer, Renaissance co-CEO — quoted in The Man Who Solved the Market, Gregory Zuckerman (2019).
Note who said it. Social media attributes this line to Simons constantly; the record has it as Mercer. We print the record.
Simons described the raw material the same way in his own words, years earlier:
Efficient market theory is correct in that there are no gross inefficiencies. But we look at anomalies that may be small in size and brief in time. We make our forecast. Then, shortly thereafter, we reevaluate the situation and revise our forecast and our portfolio. We do this all day long.Jim Simons — to the Greenwich Roundtable, quoted by Hal Lux, Institutional Investor, November 2000.
So the whole record rests on a coin landing your way 50.75 times in 100, flipped relentlessly, with the models never overruled — “we don’t override the models,” Simons told the same interviewer. That is expectancy in industrial form, and it is the opposite of what beginners chase. Nobody at Renaissance wanted a high win rate or a big call. The Stage 9 lesson on risk-to-reward and expectancy works through why a tiny per-trade edge is the only kind that survives contact with reality.
The one losing year, and what he did about it
Medallion made 8.8% in its first partial year, then started bleeding. Hal Lux’s 2000 profile puts it plainly: the fund “lost money steadily in 1989 until Simons halted trading in June.” The year closed at −4.1% — the only losing year in its history. For six months he and Henry Laufer traded nothing and rebuilt the method, replacing the forecast-the-Fed approach he had carried over from his earlier fund with a statistical one.
The loss was small enough to come back from: −4.1% needs only +4.3% to undo, and Lux reports the fund never returned less than 21% in a year for the rest of the decade. Compare that with how most accounts die — down 50% needs +100%, down 80% needs +400%. He switched the machine off while switching it off was still cheap. That is the whole content of avoiding the big loss, demonstrated by a man with 140 PhDs on staff.
The ceiling he named himself
In November 2000, asked whether there was a size limit for a firm like his, Simons answered with a number:
For years people have asked me, “How much money can you manage?” And my honest answer has been, “About twice as much as we now manage.” … We now manage a little less than $4 billion. Can we manage $7 billion or $8 billion? Yes. Could we manage $70 billion? Of course not. I wouldn’t have a clue as to how to manage that.Jim Simons — interviewed by Hal Lux, Institutional Investor, November 2000.
In the same interview he noted that his original anomalies “have weakened some” and that long commodity trends “don’t really exist anymore.” An edge is a wasting asset, and he said so.
Medallion itself stayed disciplined: closed to new money in 1993, closed to non-employees in 2005. But Renaissance built separate funds for outsiders, and those grew. Institutional Investor reported their sizes at the start of 2020 — RIEF nearly $36bn, RIDA $15bn, RIDGE $14.3bn. Add them and you get $65.3 billion: 93% of the figure Simons had called inconceivable.
In fairness, those are different funds running a slower strategy, not Medallion scaled up — that was the point of building them. But the year the three came closest to his stated ceiling, 2020, was the worst year any of them ever had.
What outsiders actually got
RIEF launched in July 2005 as the way for institutions to own a piece of Renaissance: a long-biased quantitative strategy aiming to beat the S&P 500 at a beta of 0.4 or lower, running roughly 2.5× leverage.
Here is the record it compiled. RIEF fell 16% in 2008 and another 6.17% in 2009; its longest drawdown, May 2007 to April 2009, was −35.73%, which needs +55.6% to undo. Then 2020 came: Medallion rose 76%, and RIEF fell. By January 2021 Institutional Investor put RIEF’s annualised return since launch at 8.05%, against 9.6% for the S&P 500 over the same window.
The story has a second half, and leaving it out would be dishonest: RIEF was up 22.5% through October 2024, its best year since 2011, fully recovered. But the investors who paid for the bad years were gone — RIEF had shrunk from about $36bn to under $20bn, the merged RIDA/RIDGE fund from over $29bn to $3.6bn. They bought the reputation, sat through the drawdown, sold, and missed the recovery. That is not a quant problem; it is the ordinary emotional cycle, at institutional scale.
A 50.75% edge on a $2,000 account
This is the part that should change how you trade, so here is the arithmetic in full.
Take Mercer’s number literally: you have a genuine 50.75% edge on trades risking one unit to make one. On a $2,000 account following this site’s 1% rule, you risk $20 a trade, so with no costs your expected profit per trade is 20 × (0.5075 − 0.4925) = 30 cents.
Now add the fee. Risking $20 with a stop 1% away forces a $2,000 position. At a 0.10% taker fee per side, the round trip costs $4.00. The edge is worth 30 cents; the toll is $4.00 — more than thirteen times the edge. You need a 60% win rate just to break even.
Read those bars again, because they hold the one thing you can act on. The fee scales with the position, not with the risk. Tighten the stop and you must buy a bigger position to keep risking the same $20 — so the tighter the stop, the heavier the toll on an identical edge. Loosen it to 10% and the same $20 of risk buys a $200 position, the round trip costs 40 cents, and break-even drops to 51%.
Which is why Renaissance’s answer — trade constantly, hold briefly — is the worst thing on this page to copy. They could do it because their cost per trade was effectively nil and they had direct exchange links before almost anyone. Yours is 20% of what you are risking. The honest retail version of a small edge is the opposite of theirs: fewer trades, wider stops, smaller positions. Run your own numbers in the position size calculator, then compare the break-even you get with the win rate your journal actually shows.
PRACTICE CORNER
The table above needs two numbers off your own exchange: your taker fee tier, and the smallest position you are allowed to open. Both sit on the public fee schedule of any major venue, and together they decide whether a small edge survives on your account at all. Look them up before your next trade, not after it.
These are affiliate links: we earn a commission if you sign up, it costs you nothing, and it does not change what we write. See our disclosure and the full exchange comparison.
Where this lesson stops working
50.75% is not a target you can adopt. It is the measured output of one machine on specific instruments, not a setting. Assuming an edge you have not measured is the most expensive way to misread this page.
Wider stops are not free. The table holds risk fixed at $20 and lets stop distance move — the only correct way to read it. Widen the stop while keeping the position the same and you have simply increased your risk. Position sizing is what makes a wider stop safe; without it this advice is dangerous.
The reported numbers are reported numbers. Medallion files nothing, and RIEF’s 8.05% was a snapshot taken at a low point. Any annualised figure carries the date it was measured on, and a page quoting one without that date is selling you something.
Common mistakes when learning from Simons
Treating the record as a product. Whenever a strategy is both famous and available, check which of the two Renaissance numbers you are being offered. Copying the turnover without the cost base. Trade-constantly logic only works if trading is nearly free. Hearing “quant” and thinking “safe.” RIEF ran about 2.5× leverage and still spent two years down 35.73%; a model does not remove volatility, it only picks what you are exposed to. Assuming the edge lasts. Simons said in 2000 that his original anomalies had already weakened and long commodity trends were gone — whatever works for you now has an expiry date too.
FAQ
What were Jim Simons’ Medallion fund returns?
Institutional Investor and the Wall Street Journal both report roughly 66% a year gross of fees and 39% net from 1988 to 2021, figures first pieced together by Gregory Zuckerman. They are not audited filings - Medallion publishes nothing. The gap between the two means fees took about 41% of everything the fund earned.
Could ordinary investors ever buy the Medallion fund?
No. Medallion closed to new investors in 1993 and to outside investors entirely in 2005; since then it has been open only to Renaissance partners and employees. What outsiders could buy were separate funds running a slower strategy, chiefly the Renaissance Institutional Equities Fund, or RIEF.
How did the Renaissance funds open to outsiders perform?
Badly for a long stretch. Institutional Investor reported in January 2021 that RIEF had annualised 8.05% since its July 2005 launch, against 9.6% for the S&P 500, after a 2020 in which it fell while Medallion rose 76%. RIEF later recovered, up 22.5% through October 2024, but its assets had already dropped from about $36 billion to under $20 billion.
How big was the edge behind Renaissance’s returns?
Very small. Robert Mercer, a Renaissance co-chief executive, is quoted saying the firm was right about 50.75% of the time. Simons described his targets as anomalies small in size and brief in time, re-evaluated all day long. The edge was never large; the repetition and the near-zero cost per trade were.