Following the Giants: A Guide to 13F Institutional Tracking

The Signal

Every quarter, thousands of the world's most sophisticated investors are legally required to show their cards. SEC Form 13F — filed by any institution managing more than $100 million in U.S. equities — creates one of the most valuable public datasets in finance. At any given time, over 10,000 institutional investors including hedge funds, mutual funds, pension funds, endowments, and family offices must disclose their long equity positions within 45 days of each quarter-end.

This transparency creates an extraordinary opportunity for retail and systematic investors: the chance to peer inside the portfolios of managers who spend millions of dollars annually on research, have direct access to company management teams, and have track records measured in decades. While the disclosure lag means you are never seeing real-time positions, the data still contains a remarkable amount of actionable information — particularly when you know where to look.

The key insight, validated by decades of academic research, is that 13F data is not equally informative across all positions. The signal quality varies dramatically depending on which stocks you focus on and which managers you follow. The art of institutional tracking is learning to filter noise from signal — to find the handful of positions that truly represent a manager's deepest conviction, rather than the hundreds of positions that simply reflect passive exposure or window dressing.

Why It Predicts Returns

The academic foundation for 13F-based investing rests on a counterintuitive finding. For decades, the conventional wisdom held that active fund managers, on average, fail to beat the market. This is largely true. Yet buried within these underperforming portfolios is a subset of positions where managers are genuinely skilled.

Cohen, Polk, and Silli's 2010 working paper "Best Ideas" is the cornerstone of this field. Analyzing thousands of fund managers over multiple decades, they reported that a fund's single best idea — typically its largest position by weight relative to the benchmark — historically outperformed the benchmark by roughly 1.6 to 2.1% per quarter in their sample. The same paper found that a manager's next-highest-conviction ideas also generated excess returns, but the effect diluted rapidly further down the position list, with the lower-weighted tail showing little evidence of skill. Read this as a historical, sample-specific result from an unpublished working paper — not a guarantee of future returns.

The explanation is elegant: fund managers face career and business pressures that force over-diversification. A manager who genuinely believes in only 10 stocks is forced to hold 50 or 100 to satisfy institutional mandates, track error constraints, and risk management requirements. The "real" portfolio — the one reflecting genuine conviction — is hidden inside the larger one. 13F data lets you extract it.

Verbeek and Wang's 2013 paper "Better than the Original? The Relative Success of Copycat Funds" (Journal of Banking & Finance) studied hypothetical "copycat" portfolios that replicate funds' disclosed holdings. Their finding is more nuanced than a blanket outperformance claim: on average, copycat portfolios delivered performance broadly comparable to their target funds after trading costs and expenses, and free-riding specifically on the disclosures of past winning funds did significantly better than most actively managed mutual funds net of costs. Keeping pace at all is notable, since copycats pay no management or performance fees. They also documented that this relative success increased after the SEC moved to quarterly disclosure in 2004, consistent with more frequent data improving signal quality.

A more recent working paper by Schroeder and Posch (2024), "Outperforming the Market: Portfolio Strategy Cloning from SEC 13F Filings," examined more than 150,000 cloned portfolios over 2013–2023. It reported that clones tracking top-quartile funds mirrored the originals and, in the backtest, ranked favorably against the S&P 500 on risk-adjusted measures. As an unpublished, in-sample backtest the specific magnitudes warrant caution — but the direction matches the broader literature that institutional positioning data carries a persistent signal.

Researchers have also applied machine learning to the same data. Fleiss and colleagues' SSRN working paper on deep reinforcement learning with SEC 13F holdings reported promising historical backtests, while cautioning that the training window was short and real-world performance remained uncertain — a reminder that backtested figures are not audited track records.

What the Research Actually Shows

• Fund managers' single Best Ideas — their highest-conviction positions relative to the benchmark — historically outperformed those benchmarks in Cohen, Polk & Silli's sample (~1.6–2.1% per quarter in that working paper)
• Copycat portfolios built from disclosed holdings performed comparably to the funds they tracked after costs, and free-riding on past winners beat most active mutual funds net of costs (Verbeek & Wang, 2013)
• Clone strategies tracking top-quartile funds ranked favorably vs the S&P 500 on risk-adjusted measures in Schroeder & Posch's (2024) backtest — an in-sample, unpublished result to read cautiously
• Machine-learning approaches to 13F data have shown promise in historical backtests, which authors caution are not audited track records (Fleiss et al., SSRN)
• 10,000+ institutions file 13F quarterly, creating a vast searchable dataset
• 45-day maximum lag from quarter-end to public disclosure
• The signal is generally considered more reliable when several independent top managers hold the same position, rather than a single filer
• Quarterly 13F disclosure (mandated 2004) provides more frequent data than the prior semi-annual regime, which the literature associates with stronger signal quality
• The Best Ideas effect is most relevant for managers running concentrated portfolios, where a large relative weight genuinely reflects conviction rather than index-tracking

How Meridian Uses This Signal

Meridian's institutional tracking engine monitors approximately 300 high-conviction fund managers selected for consistent long-term outperformance, concentrated portfolios, and low turnover. For each manager, we identify their "Best Ideas" — positions where their conviction significantly exceeds their benchmark weight — and aggregate these into a conviction-weighted score. When multiple top managers hold the same position, the score compounds: several independent research teams reaching the same conclusion carries more information than one.

The signal feeds into Meridian's composite Smart Money Score, which combines institutional positioning with insider buying, congressional trading, and dark pool activity. A stock scoring highly across all four dimensions has had its investment case validated by multiple independent categories of sophisticated capital. Importantly, Meridian applies a conviction filter rather than simply tracking all 13F holdings: we specifically weight against mega-cap positions (where nearly every fund holds Apple, Microsoft, and Nvidia for index-tracking reasons) and instead focus on mid-cap positions where active judgment genuinely dominates passive flows. The 45-day lag is a genuine limitation, which is why we weight 13F signals most heavily for low-turnover value and quality managers, whose positions tend to be stable across multiple quarters.

Key Takeaways

• 13F filings cover 10,000+ institutions managing $100M+ in U.S. equities, filed within 45 days of quarter-end
• Academic work (Cohen, Polk & Silli) finds fund managers' single Best Ideas historically outperformed their own benchmarks — a sample-specific working-paper result, not a forward guarantee
• Cloning research (Verbeek & Wang; Schroeder & Posch) suggests replicating disclosed holdings can keep pace with or track the originals, with the strongest evidence for concentrated, top-quartile funds
• The signal is most useful when filtered for high conviction (largest positions by relative weight) and manager quality
• Overlap among several independent top managers is generally viewed as more informative than a single filer's position
• The 45-day lag is most manageable for low-turnover strategies; pair with other, more timely signals for higher-turnover ideas
• Free to access via SEC EDGAR; platforms like WhaleWisdom and Dataroma aggregate and rank institutional positions

Expert Perspectives

"You don't need to do your own research. You need to identify who is doing first-rate research — and then follow their highest-conviction judgments." — Inspired by the Buffett approach to capital allocation, as articulated in the 13F cloning literature

Warren Buffett's Berkshire Hathaway is one of the most closely watched 13F filers in the world, with every quarterly disclosure triggering extensive media coverage. Buffett has long argued that concentrated, high-conviction investing is the path to superior long-term returns. profile →

"Diversifying well is the most important thing you need to do in order to invest well." — Ray Dalio

Ray Dalio's approach at Bridgewater emphasizes understanding why a position makes sense within a broader macro framework — not just copying positions. When tracking institutional 13F data, understanding the investment thesis behind a position matters as much as knowing the position exists. profile →

"If you avoid the losers, the winners take care of themselves." — Howard Marks

Howard Marks' second-level thinking applies directly to 13F analysis: the question isn't just what top managers are buying, but whether the information is already priced into the market by the time you act on it. The 45-day lag is real — which is why filtering for low-turnover, high-conviction managers is essential. profile →

Further Reading

• Cohen, L., Polk, C., & Silli, B. (2010). "Best Ideas." SSRN Working Paper. Key finding: in the authors' sample, fund managers' single highest-conviction holdings outperformed their benchmarks by roughly 1.6–2.1% per quarter.
• Verbeek, M., & Wang, Y. (2013). "Better than the Original? The Relative Success of Copycat Funds." Journal of Banking & Finance, 37(9), 3454–3471. Key finding: copycat portfolios performed broadly comparably to their target funds after costs, and free-riding on past winners beat most active mutual funds net of costs; relative success rose after quarterly disclosure began in 2004.
• Schroeder, J., & Posch, P. (2024). "Outperforming the Market: Portfolio Strategy Cloning from SEC 13F Filings." SSRN Working Paper. Key finding: across 150,000+ cloned portfolios (2013–2023), clones of top-quartile funds tracked the originals and, in-sample, ranked favorably against the S&P 500 on risk-adjusted measures.

Practical Tools for Individual Investors

The democratization of 13F data over the past decade has made institutional tracking accessible to any investor willing to spend an afternoon learning a few platforms. Dataroma.com is the simplest starting point: it aggregates the reported holdings of approximately 70 of the most widely followed superinvestors and hedge funds, updated after each quarterly filing deadline. You can see at a glance which stocks are most widely held, which positions were added or increased most recently, and which managers share overlapping views on a given company.

WhaleWisdom offers a more quantitative approach, providing historical 13F data with backtesting capabilities that allow you to validate how specific clone strategies would have performed over prior years. Their "Whalescore" metric ranks funds by historical stock-picking performance, giving you a pre-filtered list of the managers whose disclosures are most worth tracking. For the technically inclined, the SEC's own EDGAR database provides raw 13F XML files accessible via API, making it possible to build custom aggregation and scoring tools in Python with relatively modest effort.

Quiver Quantitative has emerged as one of the most useful platforms for institutional data in a structured, API-accessible format — particularly valuable for quantitative investors who want to incorporate 13F signals into systematic models. The key lesson from practitioners who have used these tools extensively: the quality of your filter matters far more than the breadth of coverage. Tracking 20 high-conviction managers with documented long-term outperformance is more valuable than aggregating signals from all 10,000+ 13F filers, most of whom add no incremental information beyond what passive indices already reflect.