Treynor Ratio: How AQR, Bridgewater, and Medallion Engineer Beta for Alpha

By Navnoor Bawa · YouTube: The Mathematical Trader

As quant strategies dominate 2026 allocator flows — Goldman Sachs prime services data shows a net 23% of allocators increasing quant exposure — BNP Paribas research reveals the trailing 1-year correlation between hedge fund returns and the MSCI World has reached 0.92, versus a 5-year average of 0.76. The funds that escaped that correlation trap did so by engineering the denominator of a 1965 formula. Here is the documented evidence, from SEC filings to Senate testimony.

The Geometry Most Practitioners Skip

Jack L. Treynor published “How to Rate Management of Investment Funds” in the Harvard Business Review, Vol. 43, №1 (January–February 1965). The core contribution was a characteristic line for fund performance — what became the formula (Rp − Rf) / βp. But the geometric consequence that practitioners routinely skip is this: the slope of the Security Market Line is exactly the Treynor ratio of the market portfolio, because β_M = 1, so the market's own Treynor ratio = (R_m − R_f) / 1 = R_m − R_f. Wikipedia's Security Market Line article states this precisely: "All of the portfolios on the SML have the same Treynor ratio as does the market portfolio... the slope of the SML is the Treynor ratio of the market portfolio." Therefore, a stock-picking rule follows directly: buy assets whose Treynor ratio exceeds the market's, sell those whose ratio falls below. Any asset sitting above the SML earns more per unit of undiversifiable risk than the market — and the vertical distance between the asset and the SML is Jensen's alpha. These are not two separate metrics; they are dual views of the same geometric relationship.

That construction principle creates two and only two levers for a portfolio manager: grow the numerator (generate alpha through skill) or compress the denominator (reduce net systematic risk through hedging and market-neutral construction). The documented evidence below shows each path pursued at industrial scale.

The 50-Year Empirical Anomaly That Makes This Tradeable

The entire Treynor-ratio strategy rests on a 50-year-old empirical finding: the SML is flat. CAPM predicts that higher beta should yield proportionally higher returns. It does not.

The first rigorous test appeared in Black, Jensen & Scholes (1972), “The Capital Asset Pricing Model: Some Empirical Tests,” published in Studies in the Theory of Capital Markets (Praeger). Using CRSP monthly returns from 1926–1966, they documented that the empirical SML was systematically flatter than CAPM predicted — low-beta portfolios earned more than their betas implied, and high-beta portfolios earned less. Fisher Black, in a companion paper published the same year — “Capital Market Equilibrium with Restricted Borrowing,” Journal of Business, Vol. 45, №3 (1972) — provided the mechanism: investors who cannot borrow at the risk-free rate overweight high-beta stocks instead of levering low-beta ones. This bids up high-beta asset prices, compressing their forward Treynor ratios. Low-beta assets sit persistently above the equilibrium SML.

Two decades of subsequent evidence confirmed and sharpened the finding. Fama and French (1992), “The Cross-Section of Expected Stock Returns,” Journal of Finance, 47(2), studying the 1963–1990 period, found — as stated in the paper’s abstract — that “the relation between market β and average return is flat.” Table V of the paper, which presents double-sorted average returns controlling for size, confirms that average returns are flat or decline slightly as post-ranking betas increase once size is held constant. Market beta was unpriced in the cross-section of stock returns.

This is not a mathematical curiosity. It means that investors holding high-beta portfolios are not being compensated for their incremental systematic risk. Their Treynor ratios are structurally below equilibrium. Investors who can lever low-beta positions have a structural advantage. The three case studies below document exactly how that advantage has been harvested at scale.

AQR’s Betting Against Beta: The Flat SML Formalized as a Live Strategy

Andrea Frazzini (AQR) and Lasse Pedersen (NYU Stern), “Betting Against Beta,” Journal of Financial Economics, 111(1), 2014, pp. 1–25 is the most precise published translation of the flat-SML anomaly into a trade specification. The paper was first circulated as NBER Working Paper 16601.

The paper’s model and empirical results are specific. The BAB strategy goes long a leveraged portfolio of low-beta assets (levered until each leg runs at β = 1) and short a portfolio of high-beta assets (delevered to β = 1). The net book runs near-zero market beta. The return is the spread between the over-priced high-beta side and the underpriced low-beta side of the flat SML.

Documented performance statistics from the paper itself:

• U.S. equities BAB factor Sharpe ratio: 0.75 (1926–2009) — approximately twice the Sharpe ratio of the value factor over the same period and 40% higher than momentum.
• Bond BAB factor: a Sharpe ratio of 0.81 per the published JFE paper, with highly significant risk-adjusted returns documented across the full Treasury bond sample.
• Alphas decline almost monotonically from low-beta to high-beta portfolios across US equities, 20 international equity markets, Treasury bonds, corporate bonds, and futures.
• As stated explicitly by the CBS Research Portal entry for the paper: “(1) Because constrained investors bid up high-beta assets, high beta is associated with low alpha, as we find empirically for US equities, 20 international equity markets, Treasury bonds, corporate bonds, and futures.”

Note on data: the main U.S. equity return series runs from 1926 through 2009 for the Sharpe ratio calculations in the NBER working paper; the published JFE version extends equity data through 2012. Bond and futures return data were drawn from AQR Capital Management’s internal pricing systems. AQR publishes the full monthly dataset for public replication.

In a June 2025 Bloomberg Open Interest interview (transcript via iHeart), AQR founder and CIO Cliff Asness described the live book structure: “We are quants, which means we spread our bets out fairly wide… We probably have about two thousand longs and two thousand shorts, balanced by country and mostly but not entirely balanced by industry.” That country-and-industry balance is the live implementation of beta-denominator management — the construction step that converts the Treynor ratio from a ranking tool into a building principle.

The firm’s SEC Form ADV Part 2A Brochure confirms the systematic infrastructure: “AQR specializes in quantitative investment analysis, which relies on proprietary models, utilizing a broad set of signals to generate views on investments and applying them in a systematic process.” Per AQR’s 2024 SEC Form 485APOS filing, strategies range “from aggressive, high volatility and market-neutral alternative strategies, to low volatility, more traditional benchmark-driven products… Investment decisions are made by the Adviser using a series of global asset allocation, arbitrage, and security selection models.”

Renaissance Medallion: A Senate Document and a Published Beta

The deepest primary-source evidence for the Treynor denominator being engineered rather than accepted comes from two independent sources on the Medallion Fund: a published academic analysis of disclosed returns, and a U.S. Senate regulatory investigation of the fund’s live execution infrastructure.

The Academic Record

Bradford Cornell, “Medallion Fund: The Ultimate Counterexample” (February 2020), Cornell Capital Group analyzed return data disclosed in Gregory Zuckerman’s 2019 book. The findings:

• Compound annual return, 1988–2018: 63.3%. $100 grew to $398.7 million.
• Arithmetic mean annual return: 66.1% around a standard deviation of 31.7%. Sharpe ratio exceeds 2.0.
• Beta vs. CRSP market index: approximately −1.0 (regression of Medallion’s excess returns on the market index).
• Fama-French three-factor regression: loadings on SMB and HML also negative.
• Cornell’s conclusion: “Whatever the source of Medallion’s returns, it is not a reward for risk bearing.”

This is the Treynor ratio’s domain failure made visible: a positive numerator divided by a negative denominator produces a negative ratio — which would rank the greatest hedge fund in documented financial history as the worst in any peer comparison. The breakdown is informative: it reveals that the metric was designed for portfolios that take on risk in the conventional direction, not portfolios that extract alpha while hedging market exposure to zero or below.

The Senate Record

The execution infrastructure behind Medallion’s near-zero beta is documented in a 93-page government report. On July 22, 2014, the U.S. Senate Permanent Subcommittee on Investigations held hearings on “Abuse of Structured Financial Products: Misusing Basket Options to Avoid Taxes and Leverage Limits.” The full Senate report documents how Renaissance Technologies routed more than $34 billion through basket option structures with Deutsche Bank and Barclays.

Senator John McCain’s opening statement confirmed the precise timeframe and scale: “Between the years 2000 and 2014, Renaissance exercised 60 long-term basket options with Deutsche Bank and Barclays, earning in the neighborhood of $34 billion in pre-tax profits.”

Senator Carl Levin’s opening statement explained the leverage dimension: margin rules “essentially prohibit U.S. broker-dealers from lending more than $1 to brokerage clients for each $1 of the client’s own money.” The basket options structure, documented in the Barclays testimony by Marty Malloy, worked as follows: “Barclays sells to a Renaissance subsidiary (Badger) a series of cash-settled options, which reference a basket of securities held by a wholly-owned subsidiary of Barclays (Palomino). The basket was funded by Barclays from the funds received from Renaissance as the premium for the options, plus the leverage financing provided to Palomino by Barclays. Renaissance determines the composition of the basket and the overall investment strategy.”

This explains, under sworn government testimony, how Medallion achieved effective leverage ratios beyond normal margin limits while maintaining legal structures that classified gains as long-term capital. The near-zero-beta output was the emergent result of thousands of simultaneous long/short pairs running at extreme leverage — not a declared target, but an architectural consequence documented in congressional proceedings.

Bridgewater’s Separation Architecture: Form ADV Language and Live Data

Bridgewater Associates has built the Treynor numerator-denominator separation into its formal product lineup. Where AQR harvests beta mispricing through long-short factor construction and Medallion achieves near-zero beta through statistical arbitrage at scale, Bridgewater’s approach is the most explicit: it formally registered the separation of alpha and beta as two distinct, named products.

Per Bridgewater’s Form ADV Part 2A, as reproduced by the Hedge Fund Database:

“Bridgewater believes that building portfolios based on risk allocations is more effective than using capital allocations; and that investors should consider their strategic asset allocation (beta) separate from tactical moves (alpha). Bridgewater believes investors can improve their portfolios’ overall results by separately creating a well-diversified beta portfolio… and a well-diversified alpha portfolio that reduces systematic biases.”

Institutional Investor reporting on Bridgewater’s regulatory filings confirmed: “In a regulatory filing, Bridgewater describes Pure Alpha as its ‘optimal alpha,’ which seeks to generate high returns with no bias to markets or other managers. All Weather is its ‘optimal beta strategy,’ which aims to capture the risk premiums embedded in assets by balancing its exposure to the primary drivers of market volatility.”

A 2024 fund performance document for Pure Alpha Major Markets, available on Scribd, states directly: “Pure Alpha Major Markets is a global active investment strategy… designed to generate a high return-to-risk ratio through active management while being uncorrelated to markets and other managers. We seek to achieve this goal by trading a highly diversified set of liquid global markets with no bias to be long or short any market over time.”

The performance record that validates this architecture was documented in Institutional Investor’s January 2020 article, citing a Bridgewater investor document: from inception through end-2018, Pure Alpha had a 0.19 correlation with equities, 0.15 with bonds, and 0.07 with other hedge fund managers. In 2008, when the S&P 500 returned −37%, Pure Alpha gained +9.4%. That is the Treynor denominator working in real time: near-zero equity beta during the largest equity drawdown in a generation meant the market’s losses were not transmitted to the portfolio.

A former member of Bridgewater’s investment committee, Robert Elliott, described his role in an SEC Form 485APOS filing from June 2022 as having “created many of the strategies for the flagship Pure Alpha fund across equities, fixed income, credit, exchange rates, and commodities. In his role on the Investment Committee, he was holistically responsible for the Pure Alpha foreign exchange portfolio from investment strategy to trade execution.”

Most recent performance data: Reuters (December 31, 2025) reported Pure Alpha’s flagship fund gained 33% in 2025 (through December 29) — its best year in 50 years. Bloomberg (January 2, 2026) and the cited Fortune league table put the full-year figure at 34% for Pure Alpha II specifically, representing the final two trading days of the year. Reuters also reported in July 2025 that Pure Alpha 18% volatility had already posted 17% in the first half of 2025, while the S&P 500 ended June up roughly 5.5%.

Per Fortune’s full-year 2025 league table and Goldman Sachs prime services data as reported by Yahoo Finance, quant funds gained 10.5% in aggregate in 2025 — with individual standouts well above that average: AQR’s Apex multistrategy returned 19.6%, D.E. Shaw’s Composite returned 18.5% and its Oculus fund 28.2%. Quant funds accounted for more than 70% of the industry’s $78 billion in net inflows, the leading strategy for a second consecutive year.

The Treynor-Black Model: Sizing the Active Book by Signal Quality

The three case studies above document what institutional-scale beta management produces in practice. Treynor’s 1973 paper with Fischer Black — “How to Use Security Analysis to Improve Portfolio Selection,” Journal of Business (1973), vol. 46, pp. 66–86 — answers the construction question that underlies all three: given alpha forecasts on a limited number of securities, how should the active portfolio be weighted to maximize the Treynor ratio of the combined book?

The model specifies a two-component structure. The passive component is the market index portfolio. The active component is a separate portfolio where each security’s weight is proportional to its forecast alpha divided by the variance of its residual (unsystematic) risk:

w_i ∝ αᵢ / σ²(εᵢ)

The total portfolio combines this active book with the passive market index in a ratio determined by the active portfolio’s own squared Sharpe ratio. The insight: at the security level, this is Treynor-ratio logic applied per-position — alpha per unit of residual noise. A security with high predicted alpha but noisy returns gets discounted; a security with modest but precise alpha gets full weight. Signal quality, not conviction magnitude, determines allocation.

A 2024 SSRN paper extending the model derives closed-form solutions under a Fama-French multi-factor setting, showing that single-factor Treynor-Black underweights the active book when additional systematic factors are present — the active portfolio should be larger when the manager has edge in factor exposure that the passive side cannot replicate.

A 2021 University of Dayton applied study applied the model to a Morningstar database of long/short hedge funds, predicting information ratios and computing Treynor-Black weights from those predictions. Result: model-weighted allocations outperformed equal-weighted portfolios in bull markets by concentrating capital where signal-to-residual-noise ratios were highest.

Execution: What Beta Management Looks Like at the Trade Level

The Treynor denominator is not achieved — it is estimated and managed. Three documented mechanisms:

Futures overlays. A long equity book with residual β = 0.8 is reduced to β = 0.0 by shorting equity index futures in the appropriate notional. Bridgewater’s SEC-documented strategy explicitly uses liquid futures across equity, bond, commodity, and currency markets for both Pure Alpha positioning and All Weather beta replication. The execution challenge: beta estimates from trailing regressions shift as correlations change. A portfolio “hedged” to β = 0 on a 36-month regression may run β = 0.3–0.4 on a realized 1-month basis.

Beta as a capital efficiency variable. AQR’s published Insights piece “Should Hedge Funds Hedge?” argues: for an investor allocating out of equities, a market-neutral strategy with β = 0 may be less capital-efficient than a strategy with β = 1.0 delivering the same alpha. Adding beta back via a zero-cost futures overlay preserves full equity exposure while capturing the alpha. The investor’s total Treynor ratio becomes: alpha / beta-from-overlay. Maximizing the ratio, in this construction, means maximizing alpha — the beta is held constant architecturally.

The 37% systematic risk floor in BAB. A Lancaster University research paper presented at the 2018 Finance of Finance conference, examining BAB factor risk, found that “on average 37% of the risk of BAB is systematic — so the market neutral strategy has a substantial market risk component.” Even a book explicitly constructed for beta neutrality retains material systematic exposure. The Treynor denominator is not zero simply because the construction targets zero — execution imprecision, intraday correlation shifts, and factor model misspecification create residual systematic exposure that shows up as beta in any post-hoc regression.

Where the Ratio Breaks: The Published Evidence

The negative beta case. A November 2025 paper in the Journal of Asset Management (Springer) formally demonstrates the anomaly: “there exists an anomaly when betas are negative. In case of the negative beta, the lower the excess return the higher is the Treynor Ratio.” The paper proposes a Modified Treynor Ratio (MTR) that preserves monotonicity. Without such a correction, as the Medallion case proves, the best-performing fund in documented financial history ranks last on the metric that should identify it.

Critique of BAB construction. A 2022 Journal of Financial Economics paper by Novy-Marx and Velikov, Vol. 143(1), pp. 80–106 — “Betting Against Betting Against Beta” — examines the BAB factor construction’s non-standard beta estimation and equal-weighting effects, concluding that “BAB earns positive returns after accounting for transaction costs, but earns these by tilting toward profitability and investment” rather than pure leverage-constraint arbitrage. The Treynor ratio premium is real; its source is more complex than the original model specifies.

Factor model incompleteness. A single-factor beta against the S&P 500 misspecifies systematic risk for any multi-asset portfolio. A global macro fund with near-zero equity beta may carry 0.6–0.8 duration beta against global rates — substantial systematic risk invisible to equity-centric Treynor calculations. Fama and French (1992) themselves showed that beta was unpriced once size is controlled, implying the single-factor model captures only one dimension of systematic risk.

Beta instability. FinWiz’s documented analysis summarizes the practitioner reality: “Beta is calculated from historical data and can change over time. A portfolio’s beta during the measurement period may differ from its forward-looking beta, making the Treynor ratio backward-looking.” Rolling beta estimates for the same portfolio can differ by 0.3–0.5 depending on window length and whether the estimation period includes a market stress event.

The Practitioner’s Decision Framework

The Treynor ratio is analytically useful in exactly the scenario where its assumptions hold: a well-diversified portfolio, positive and stable beta against a relevant benchmark, and a measurement period without regime change. Outside those conditions, it produces rankings that range from misleading to actively wrong.

For a quant PM building a book:

• If the portfolio is diversified and β_p > 0: compute the ratio and compare against the market’s Treynor ratio (R_m − R_f). Anything above that threshold has Jensen’s alpha > 0. Size the active book using Treynor-Black weights — proportional to alpha-per-unit-of-residual-noise, not conviction magnitude.
• If β_p ≈ 0 (market-neutral): the denominator is near zero and the ratio is uninformative. Evaluate on Jensen’s alpha or the information ratio instead. Use the Treynor ratio only to confirm the hedge is working — a ratio that explodes upward confirms near-zero realized beta, but tells you nothing about alpha quality.
• If β_p < 0: the original Treynor ratio produces inversely-ranked results. Apply the Modified Treynor Ratio from the 2025 Journal of Asset Management paper, or evaluate on absolute returns and drawdown during market stress events — the 2008 and March 2020 performance of a negatively-correlated book is more informative than any calm-period ratio.

The through-line across AQR’s Betting Against Beta, Renaissance Medallion’s basket-option-leveraged stat arb (documented under Senate testimony), and Bridgewater’s formally registered alpha-beta separation is structural: each fund identified a regime where the Treynor ratio’s denominator was mispriced or architecturally controllable, and built portfolio construction around that insight. The SML was not a benchmark — it was a map of where systematic risk was structurally overpriced, and compressing the denominator was the mechanism that populated the numerator.

📊 Want Deeper Quantitative Analysis?

This research took a very long time of data collection, verification, and analysis — primary sources tracked from NBER working papers to Senate testimony transcripts. If you found value in this deep-dive, I publish exclusive quantitative research, trading strategies, and institutional-grade analysis on Patreon.

By joining, you’ll be supporting my work and motivating me to publish more content like this.

**→ Join the Patreon community here**

Follow along for more:

• 📺 YouTube — The Mathematical Trader
• 💼 LinkedIn — Navnoor Bawa

Primary sources: Treynor (1965), HBR Vol. 43 №1 · Black (1972), JBus 45(3) · Black, Jensen & Scholes (1972), SSRN 908569 · Fama & French (1992), JF 47(2) · Frazzini & Pedersen (2014), JFE 111(1) · AQR BAB Monthly Dataset · AQR Form ADV · AQR Form 485APOS (2024) · Cliff Asness Bloomberg interview transcript · Cornell (2020), Medallion Counterexample · Senate PSI Hearing July 22, 2014 · Senate PSI Full Report (93pp) · Barclays Senate Testimony (Malloy) · McCain Opening Statement, PSI (2014) · Levin Opening Statement, PSI (2014) · Bridgewater Form ADV (HedgeFundDB) · Bridgewater Pure Alpha Major Markets Document (Scribd, 2024) · Bridgewater Elliott SEC Filing (2022) · Institutional Investor: Bridgewater 2008 performance · Institutional Investor: Bridgewater regulatory filing · Reuters: Pure Alpha 33% full year 2025 · Bloomberg: Pure Alpha II 34% 2025 · Fortune: 2025 hedge fund returns league table · Reuters via TradingView: Pure Alpha H1 2025 · Goldman Sachs prime services data via Yahoo Finance (Feb 2026) · Goldman Sachs: Hedge Funds Have Momentum (Feb 2026) · BNP Paribas 2026 Hedge Fund Outlook · AQR: Should Hedge Funds Hedge? · Modified Treynor Ratio, J. Asset Mgmt., Springer (2025) · Treynor-Black Model, Wikipedia · Lancaster/FoFi 2018: BAB systematic risk · Novy-Marx & Velikov, JFE Vol. 143(1) (2022)