Key takeaways
A real edge decays because the moment it works, competitors find it.
He stopped on purpose, closing a winning fund when the math changed.
Track records hide the mechanism. Ask what generates returns, not just how big.
Fee models reward staying even after the strategy has stopped paying off.
The most studied track record in hedge fund history
In 1988, Edward Thorp closed a hedge fund that had never had a losing year. The trade still worked. Competitors still couldn't replicate it. He shut it down anyway, and within a few years the strategy he had run was effectively dead across the industry.
That sequence runs backward to how managers behave today. Thorp usually gets told as a biography. Card counter beats the casino, takes the math to Wall Street, walks away rich. The casino lore is the least useful part. The edge at Princeton-Newport Partners was narrow and technical, and understanding how it worked, and why it expired, tells you more about underwriting a manager's "structural advantage" claim in 2026 than any blackjack story does.
The numbers with important caveats
Per Schwager's Hedge Fund Market Wizards, which documents the fund's full record:
~15.1% annualized net of fees over 19 years (~19.1% gross)
227 winning months, 3 losing months, none worse than 1%
A 98.7% winning-month record across 230 months
Zero losing years, November 1969 to December 1988
Jay Regan, Thorp's partner, told Morgan Housel in 2017: "We were in business for 22 years and never had a losing quarter."
Asked whether two young men could replicate that record today, Regan was direct: "No. It's too competitive now."
The trade was narrower, more boring, and more replicable-looking than the legend suggests
What Princeton-Newport actually did, in plain terms:
Buy a convertible bond when its embedded equity option was underpriced
Short the underlying stock in a calculated ratio
Capture the spread
Hedge the directional risk
Market-neutral by construction. Whether the broader market rose or fell was irrelevant to whether the trade paid.
The edge in 1969 was not the trade design. Variants had existed for a decade. The edge was that perhaps a dozen people on Wall Street could price a convertible accurately. Thorp and Sheen Kassouf had published Beat the Market in 1967, six years before Fischer Black, Myron Scholes, and Robert Merton formalized the closed-form option-pricing solution that became the industry standard. Thorp's approach was empirical rather than identical, but the practical effect meant he could price an instrument that his counterparties could not price back.
A structural edge is a technical answer to the question of what a manager knows that the market cannot easily replicate. Fundraising materials today tend to substitute team biography, sourcing networks, or process language for that answer. Thorp's version was different in kind. He could point to a formula no one else had.
Kelly sizing, and why it does not survive modern fee math
Thorp sized every position using fractional Kelly logic, typically half the full Kelly bet to keep volatility in check. The Kelly criterion, developed at Bell Labs in 1956, is a formula that says, given a known edge, exactly how much capital to put behind a bet to compound as fast as possible without risking ruin.
Bigger edge, bigger position
Smaller edge, smaller position
Zero edge, zero position
Two consequences for any manager claiming a comparable approach:
Kelly only works if the manager genuinely knows the edge and can quantify it. Estimation error compounds quickly. Most managers do not have the inputs Kelly assumes.
Kelly caps deployable capital at the level the edge supports. A manager earning two-and-twenty on a multi-billion-dollar book cannot honestly operate that way. The management fee is calculated on assets, not on alpha, so the economics demand putting more capital to work than the edge justifies.
That second point is where the historical case becomes a present-day diligence problem. When a strategy scales past its honest capacity, returns typically dilute, factor risks creep in, or both. Allocators see the symptoms in degrading Sharpe ratios. The cause rarely makes it onto the deck.
The fee model that aligned Thorp with his LPs is functionally extinct
Princeton-Newport charged a performance fee with effectively no separate management fee during much of its early operation. The economics forced alignment. Thorp ate only when the LPs ate. There was no AUM revenue floor to subsidize a strategy that had stopped working.
The same structure today would fail most institutional onboarding, prime brokerage diligence, and fund administration setups, all of which presume a management fee sufficient to fund a compliance and operations stack that did not exist in 1969. So the structure is not a recommendation. It is a benchmark for what alignment looks like when nothing else is propping the manager up.
The modern fee model creates a survival incentive independent of edge. The manager continues to earn on AUM whether or not the strategy is still paying. That incentive is doing real work in why funds stay open past the point their edge has decayed, and it is the unstated counterweight to every conversation about private markets fee compression.
When LPs negotiate management fees down 10 or 20 basis points, directional alignment improves at the margin. The underlying geometry does not.
Objection from a private-markets allocator, and where it falls short
A reasonable pushback runs as follows. Convertible arbitrage was a pricing edge in a public, observable market. Once the math became commoditized, the trade compressed visibly. Private markets work differently:
A PE firm's edge can be sourcing-based, built from operator networks over a decade.
A real estate manager's edge can be sub-market specialization that takes a fund cycle to develop.
A private credit manager's edge can be structuring expertise concentrated in a few originators.
None of these compress as fast or as visibly as a public-market pricing edge does. The Thorp framework, the argument runs, does not generalize. The decay clock runs more slowly in private markets and is harder to observe in real time. There is no exchange-traded equivalent of a convertible mispricing that compresses on a six-month curve.
The decay still happens. It just runs slower, which gives the manager more room to deny what is happening to anyone underwriting them:
Sourcing networks tend to saturate as competitors hire from the same operator pool. The mid-market healthcare buyout segment between 2014 and 2022 is a reasonable reference for how visibly that can play out.
Operational playbooks tend to standardize as consultants productize them across portfolios.
Sector specialization erodes as capital floods the segment.
The implication for allocators is that a private markets GP should be asked what a quant manager would be asked. What is the specific mechanism by which this edge erodes? Where on that curve is the fund? What would the manager have to see to return capital?
How Thorp read his own decay curve, and acted on it
By the mid-1980s, the spread that funded Princeton-Newport's first 15 years had compressed:
Drexel Burnham Lambert had industrialized convertible issuance
A generation of Black-Scholes-trained quants were running similar books at investment banks
The trade was no longer scarce
Princeton-Newport itself closed in 1988-89 under financial pressure from a RICO investigation related to the Drexel circle, not because Thorp had concluded the edge was finished. Five Princeton-Newport executives and a former Drexel trader were convicted in 1989 on counts including racketeering, securities fraud, and tax fraud. The convictions were largely overturned on appeal in 1991. Thorp was never indicted.
The more useful behavioral evidence came later. He launched Ridgeline Partners in August 1994 on a statistical arbitrage strategy, a different inefficiency entirely. Ridgeline ran for eight years at roughly 21% annualized. He wound it down in 2002 as more funds adopted the same approach and the returns it depended on compressed.
Thorp's later remark in Schwager's Hedge Fund Market Wizards captured the broader shift: "The hedge fund industry is following the path of the mutual fund industry, which became equivalent to the market, or actually worse than the market once you account for costs."
Not a comment on hedge funds in particular. A comment on what happens to any asset class once fee economics outlast edge.
The fund that saw the same decay and made the opposite call
Thorp's instinct looks obvious in hindsight. The cleaner test is what a comparable fund did when handed the same signal.
Long-Term Capital Management launched in 1994 on convergence trades, the same market-neutral premise Thorp had used: identify two securities whose prices should move together, bet on the spread closing, stay neutral to market direction.
The pedigree was extraordinary. John Meriwether's bond-arbitrage team from Salomon, plus Myron Scholes and Robert Merton, who shared the 1997 Nobel for the option-pricing work that sits underneath both LTCM's trades and Thorp's. Returns ran at 40% in 1995 and 1996.
Then the edge began compressing, for the same reason Thorp's had. Competitors copied the trades, and the spreads LTCM lived on narrowed. The fund saw it happen.
At the end of 1997, LTCM returned about $2.7 billion to investors, cutting its capital base roughly 36% to $4.8 billion.
Shrinking the capital while holding the positions did one thing mechanically: it pushed leverage higher, to roughly $30 of debt for every $1 of capital.
It expanded into merger arbitrage and equity-index options, trades where its informational advantage was thinner. The move drew internal opposition from Scholes and Merton.
Handed a decaying edge, LTCM levered the remainder harder and reached for new trades it could not price better than anyone else. When Russia defaulted in August 1998, LTCM's losses compounded fast. A single month wiped out $1.8 billion, cutting the fund's capital to $2.3 billion, and on September 23 a consortium of fourteen firms put up roughly $3.6 billion to take it over.
Same era, same market-neutral premise, same Nobel-grade math, same compressing edge. Thorp returned the capital and shut down. LTCM borrowed against the shortfall. The pedigree did not determine the outcome. The willingness to stop did.
The diagnostic Thorp ran on Madoff in 1991, and what it teaches about reading a deck
An institutional client hired Thorp in 1991 to evaluate their portfolio. One of the positions was with Bernard L. Madoff Investment Securities. Thorp identified pattern inconsistencies in the reported trades within a short review window, confronted Madoff's nephew, and recommended his client liquidate. They did.
Seventeen years later, the SEC's complaint against Frank DiPascali confirmed that none of the split-strike conversion trades Madoff claimed to be executing had ever happened. Thorp could price the trades Madoff claimed to be running and saw the returns were geometrically impossible at the option volumes available on the relevant exchanges. He did not need a whistleblower. The math was enough.
The same logic, applied to a modern fundraising deck, asks four questions:
What is the specific, technical answer to "what do you know that competitors cannot replicate"?
How is position sizing set relative to the actual edge available, and what happens to it as AUM grows?
Does the fee structure permit honest assessment of decay, or punish the manager for it?
What would the manager have to observe to return capital, and have they ever done so before?
The prepared answers reveal little. The hesitations reveal where a manager has actually thought about the edge and where they have only assumed it.
Bottom line
The managers who outlast their cycle tend to share something with Thorp that does not show up on a track record summary. They have built the operating conditions that let them act on decay when they see it:
Carry-weighted compensation rather than AUM-weighted compensation
Fund size kept below the edge's honest capacity, even when the next vintage could have been raised twice as large
A communication posture with LPs that signals decay risk transparently rather than burying it under process language
A structure in which the manager's personal capital sits beside the LPs' rather than upstream of them
None of those conditions are visible in a deck. They show up in how a GP answers questions about capacity, succession, and what would cause them to wind down.
The underwriting move that compounds across cycles is shifting weight away from the historical return number and toward those four operating conditions. The next decade of distributions will be earned by the managers who built the structure to stop when stopping is the correct call. Most of the survivors of the 2030s, in private and public markets alike, will look in retrospect more like Thorp than like the managers currently raising on the back of vintages that worked.
For the managers building toward that, the harder task is making an honest account of edge decay legible to allocators who have been trained to discount it. That is a communication problem before it is a fundraising one, and it is the kind of work Collateral Partners does with private-markets firms.
