The investor library

Jim Simons

1938-2024 · Renaissance Technologies · The best record ever recorded, and the least useful.

Overview

A geometer who never took a finance course and produced the highest returns in the history of the industry. Renaissance's Medallion fund compounded at roughly 66% a year before fees between 1988 and 2018 — about 39% after a fee structure so extreme that it would be laughable attached to any other record.

Background

Newton, Massachusetts, 1938. MIT mathematics at twenty, a Berkeley doctorate at twenty-three. He broke codes for the Institute for Defense Analyses during Vietnam and was dismissed after publishing a letter opposing the war — an early demonstration that he was not employable in a normal institution.

He then chaired the mathematics department at Stony Brook for a decade and made it serious, and did the work he is respected for among mathematicians rather than investors: Chern-Simons theory, a piece of differential geometry from 1974 that turned out to matter in theoretical physics. He left academia for trading in 1978, and founded what became Renaissance in 1982.

Style, and how it evolved

He began as a discretionary macro trader and was not good at it. This is the most important fact on the page and the one most often skipped. He traded currencies on judgement, found the experience unbearable and the results erratic, and concluded the problem was him.

So he removed the human. Renaissance hired mathematicians, physicists, statisticians and — decisively — speech-recognition researchers from IBM, including Robert Mercer and Peter Brown, who arrived with the insight that markets could be modelled the way language is: enormous data, hidden states, no need for a story about why.

Very many small edges. Short holding periods, vast numbers of trades, and a win rate reportedly a fraction above even. At sufficient scale and frequency, a signal that is right 50.75% of the time is an extraordinary business. Nothing depends on any single position being right.

No narrative required. If a pattern survives rigorous out-of-sample testing it is traded, whether or not anyone can say why it exists. This is the deepest break with every other investor in this library, all of whom require a reason.

Data cleaning as the actual edge. Much of the early advantage came from assembling and correcting historical price data nobody else had bothered to make usable.

Performance

paid offa decisiona losswalked awaythe lifeshaded columns are the crashes — hover any mark

Medallion, 1988-2018: about 66% a year gross, roughly 39% net of fees ✓ — documented in Gregory Zuckerman's reporting and corroborated by figures disclosed in litigation and regulatory proceedings. The fund charges a 5% management fee and 44% of profits; at any lesser record the gap between those two numbers would be the story.

Individual years commonly cited include roughly +98% in 2000 and a very large gain in 2008, both while markets fell hard. ○

✓ documented · ○ widely reported, figures drift between accounts

Case studies

1989 — nearly shutting down. The early Medallion lost money over a sustained stretch, and Simons considered closing it. What came out of that period was the decision to stop overriding the models with judgement. The mechanism: the discipline is not that the system is smarter than the operator. It is that the operator, having been measured, agreed to stop interfering. ✓

2008 — uncorrelated, and therefore useful. Medallion returned a very large gain in a year when nearly every other strategy failed together. The signals had no economic thesis and therefore no shared exposure to the thing that broke everyone else. It is the clearest argument that the absence of a narrative is a feature. ○

2020 — the same firm, two answers. Medallion gained heavily while Renaissance's public funds — same building, same researchers, far larger capacity, longer holding periods — lost heavily enough that investors withdrew billions. The mechanism: the edge does not survive being made large and available. This is the single most instructive fact about Renaissance and it is rarely the one quoted. ✓

The other side of the record

A record this good is where scepticism is most worth spending, so:

It is unlearnable by construction. The method is secret, protected by litigation against departing employees, and reportedly not fully understood by any single person inside the firm. There is nothing to adopt.

It does not scale, and the firm's own products prove it. Capacity is capped in the low tens of billions and the strategy is returned to employees precisely because outside money would destroy it. Any strategy that must stay small is a personal fortune, not an investment approach.

The net-of-fee gap is enormous. 5% and 44% converts 66% into 39%. That the investors tolerated it is a measure of the returns; it also means the achievement being celebrated is substantially the manager's, not the client's.

A large part of the after-tax outcome came from a structure regulators disputed. The basket-options arrangement examined by a US Senate subcommittee in 2014, which converted short-term gains into long-term ones, was settled in 2021 with executives agreeing to pay up to roughly $7bn. Any honest accounting of the returns has to include the part that was argued about.

And the survivorship problem is total. Quantitative funds fail constantly and quietly. One extraordinary survivor, with an undisclosed method and no replication, is a data point about the possible — not evidence about what works.

Key lessons

• A tiny edge, repeated enormously often, beats a large edge held rarely — provided the costs are low enough and the repetitions are genuinely independent.
• Know your capacity. The most valuable strategies are usually the ones that stop working when they get big, and recognising that early is what preserved this one.
• Data quality is an edge, and an unglamorous one. Much of the early advantage was cleaning prices nobody else would.
• Hire for problem-solving, not for pedigree in the field. Renaissance's decisive hires came from speech recognition and code-breaking.
• He fired himself first. The founding act was measuring his own discretionary trading, finding it wanting, and removing it. Very few people in this library did that.

Reading and links

• The Man Who Solved the Market — Gregory Zuckerman, 2019. The only serious account, and unauthorised.
• Simons's public lectures — MIT and the Numberphile interviews. He talks freely about everything except the method.
• The 2014 US Senate Permanent Subcommittee on Investigations report on basket options — the adversarial primary source.

Marked ✓ where documented and ○ where widely reported but drifting between accounts. This is a profile of an investor, not a view on any security.

Primary sources

Go to the thing itself. These are the subject’s own publications or an institution’s own site. No bookseller links, and no referral arrangements — a reading list that earns per click is not a reading list.

• Renaissance TechnologiesMinimal by design, which is itself the point.

Profiles of investors, not views on any security, and not personal advice. Figures are marked ✓ where they are documented and ○ where they rest on secondary or private accounts — investing biography is heavily mythologised and the well-known numbers drift with each retelling. Capital is at risk.

Corrections and right of reply. Where a criticism is made of a named person it is stated as a specific measure over a stated period, attributed to its source, and separated from opinion. Assessments are opinion, honestly held, on facts believed accurate at the date shown. If anyone profiled here — or anyone acting for them — believes a fact is wrong, it will be corrected promptly and visibly, and a reply will be published alongside it on request.

← Howard MarksBill Miller →