The Record You Cannot Take With You: What Counts as a Verified Quant Track Record

In June 2026 a federal judge in Manhattan refused to dismiss a criminal trade-secrets charge against a former Headlands Technologies quantitative researcher, over trading components the firm says cost more than $1 billion to develop. Prosecution is the rare end of a problem every senior quant meets in a milder form on the way out of a good seat. The profit and loss stays on the platform's servers. What leaves with you is testimony.

A verified track record is a performance history that someone other than you can confirm: dated returns or profit and loss attributable to one named decision-maker, corroborated by the firm that booked them, a fund administrator, an auditor, or a regulator. As of September 2026, very little of what a portfolio manager carries out of a multi-manager platform meets that definition on the day they resign. The distance between what a senior candidate believes they can evidence and what a buyer will accept is the most common surprise in a senior process, and it is entirely avoidable with six months of warning.

What does a buyer actually mean by a verified track record?

Three buyers ask the question, and they mean three different things by it.

A hiring platform wants attribution it can reason about: the mandate you ran, the capital behind it, the risk framework you sat inside, and how much of the result was yours rather than the framework's. It resolves this through references rather than documents.

An allocator or seed investor backing a launch wants numbers an administrator and an auditor stand behind. The Global Investment Performance Standards exist for that purpose, and at the end of 2023 1,468 of the 1,778 organisations claiming compliance had the claim verified by a third party. GIPS also governs carrying a record across firms, and the determining question is not the old firm's compliance but whether the new firm continues the same strategy with the same continuing decision-makers.

A regulator asks the hardest version. Under SEC Rule 206(4)-1, an adviser may show predecessor performance only where the manager was primarily responsible for the results, the prior accounts were sufficiently similar to the current mandate, every similarly managed account is included, and the disclosure makes the provenance plain. The clause that removes most platform seats from consideration is the committee test: where predecessor performance came from collective decisions, most of the responsible decision-makers must join the successor firm before the record travels with anyone. SEC staff has since eased the arithmetic of presentation, updating its Marketing Rule guidance on 19 March 2025 to allow gross figures for extracted investments, and gross portfolio characteristics such as yield or Sharpe ratio, without their net equivalents in defined circumstances. The staff's latest additions, dated 15 January 2026, dealt with model fees and testimonials. None of it eased anything about attribution.

Why the numbers you remember are not the numbers you can show

Inside a platform, the record is the firm's record. The order management system, the risk overlay, the financing and borrow costs, the centralised hedging, and the allocation decisions that sized your book up and down through the year all belong to the employer, and all of them sit between your trading decisions and the number you would like to quote. A gross figure that ignores platform financing is not the figure the platform reported internally, and a buyer who has run pods knows the difference.

A letter from the firm confirming role, period, mandate and result is the clean solution, and it is entirely discretionary. Firms grant it to people who leave well and withhold it from people who leave badly, which is why the request belongs in the conversation before notice is given rather than after.

The platform has the academic record on its side when it argues the number belonged partly to the room. Groysberg, Lee and Nanda tracked star equity analysts who changed firms and found their performance fell on arrival and stayed lower for as long as five years, with the damage smallest for those who moved with their teams and largest for those who moved alone into a weaker platform. Bidwell's six years of investment-bank data pointed the same way from the buyer's side: external hires were paid around 18 percent more than internal promotions and evaluated worse for their first two years, and left at higher rates. A hiring firm that discounts an imported number is pricing that literature, whether or not anyone in the room has read it.

What is reliably verifiable is narrower and duller. FINRA's Form U4 carries ten years of employment history, inside and outside the securities industry, visible through BrokerCheck; advisers registered only with the SEC appear instead on Investment Adviser Public Disclosure. In the United Kingdom, Senior Managers and Certification Regime rules oblige a hiring firm to request a regulatory reference from a candidate's past employers going back six years, for senior managers and certification staff alike, and since the regulator's April 2026 reforms took effect the expected turnaround on that reference has been four weeks rather than six, under policy statement PS26/6. Those sources establish where you sat, for how long, and whether you behaved. Not one of them reports a single basis point of performance. The performance layer travels through people, which is why a confidential search runs on relationships rather than paperwork.

How long does a record have to be before it means anything?

Longer than most platform tenures, and the statistics are unkind about the gap. The records that clear the bar without argument tend to be fund-level and old: Quantedge, the Singapore systematic manager, was reported in August 2026 to be up 34.6 percent for the year on a record running close to 20 percent a year since its 2006 launch, which is the kind of number an allocator checks with the administrator rather than the manager. A two-year platform book is the opposite object. Andrew Lo's analysis of Sharpe ratio estimation in the Financial Analysts Journal established that monthly Sharpe ratios cannot be annualised by multiplying by the square root of twelve except under narrow conditions, and that serial correlation in monthly returns can overstate a hedge fund's annual Sharpe ratio by as much as 65 percent. Correcting for it reorders manager rankings substantially.

Selection makes it worse. Bailey and López de Prado's deflated Sharpe ratio prices the fact that the expected maximum Sharpe ratio across many trials rises with the number of trials, so a strong figure picked from a set of candidate configurations carries far less information than the same figure produced once. Their earlier work on the probabilistic Sharpe ratio formalised the minimum track record length needed for a Sharpe estimate to clear a threshold at a stated confidence, and showed that requirement lengthening as returns get more skewed and fatter tailed, which describes most of the strategies worth hiring for.

The base rate underneath all of it is the one every allocator has internalised. Barras, Scaillet and Wermers, applying a false-discovery framework to fund alphas in the Journal of Finance, found roughly 75 percent of funds sitting at zero alpha net of costs and 0.6 percent genuinely positive. A buyer who discounts a two-year platform book is not being difficult. They are applying the correct prior.

Which part of your record actually predicts anything?

The part nobody volunteers. Sun, Wang and Zheng, in Federal Reserve Board working paper FEDS 2016-030, split hedge fund returns by whether they were earned while the sector as a whole was weak or strong, and found persistence in only one of the two. Funds in the top quintile of returns earned during weak markets beat the bottom quintile by around seven percent over the following year on a risk-adjusted basis, and the signal still had predictive content three years out. Returns earned during strong markets predicted nothing.

Senior candidates lead with their best year. The evidence says the best year is close to the least informative thing in the file, and that the drawdowns carry the content: what you cut, when you cut it, what you refused to cut and why, how the book behaved in the weeks when the whole strategy was wrong. That material needs no letter from the firm and no data extract. It is a conversation, it is corroborated by the risk officer and the head of research who watched you have it at the time, and it survives every restriction in this piece intact.

What can you take, and what ends in a courtroom?

The federal answer got simpler and less generous over the past year. The Federal Trade Commission's non-compete ban never took effect, the agency voted three to one on 5 September 2025 to dismiss its appeals and accept vacatur, and the rule was removed from the Code of Federal Regulations on 12 February 2026. Restrictive covenants are a state question again. Outright bans stand in a handful of states including California, Minnesota, North Dakota and Oklahoma, and a dozen states plus the District of Columbia gate enforceability on income at thresholds no senior quant comes close to clearing: the District's 2026 figure is $162,164. Garden leave is a separate instrument, contractual and paid, and it binds where it is written.

Trade secrets mark the harder boundary, and the Headlands case marks it clearly. The charge covers proprietary components the firm calls Atoms and Alphas, described as the building blocks of its trading infrastructure; the defendant was arrested in January 2025 and faces up to ten years on a single count. Criminal exposure of that kind stays unusual in quantitative finance, where the civil suit between a firm and a departing employee is close to routine. The operative line has not moved: what you understand is yours, and what sits on a drive, in a repository, or in an exported file is theirs, whether or not you wrote it.

What to have in hand before you give notice

Four things, and every one of them is easier to obtain as a colleague than as a leaver. Confirmation from the firm, in whatever form it will give, of your mandate, your period, your capital and your result. Three or four named people, including at least one from risk or operations rather than the investment side, who watched you run the book and will say so. A written account of your own process, decision rules and limits, composed from memory rather than from any firm document. And a clean regulatory footprint, checked yourself on BrokerCheck or the adviser register before a hiring firm checks it for you, because dates that fail to reconcile are the fastest way to lose a process you were winning.

Very little of this is what a senior quant move is usually prepared for, and the reason is timing. The decision to leave a platform is normally taken in the weeks after a bonus number lands, while the evidentiary work takes months and depends on people who are still colleagues. The regulators have meanwhile shortened the other side of the clock: a London hiring firm now expects the regulatory reference on its new portfolio manager back within four weeks of asking, so the official account of your last seat is written, sent and read before most garden leaves have properly begun. The candidates who move well out of the next pay round will be the ones whose references were settled while there was still nothing to explain.