Concentrated portfolios look more alike than in 2015, and almost all of the increase comes from megacaps

Why it matters

Many investors read overlap among "famous managers" as meaningful: if several managers with concentrated portfolios hold the same stock, it would reflect shared conviction. This report answers two questions for anyone following 13F portfolios: how much of that overlap is simply common exposure to the market's largest stocks and how much real diversification you get by combining several managers' portfolios.

1. Two groups: one converges on the megacaps, the other holds none

The average number of megacaps per concentrated portfolio went from 1.70 to 2.70 between Q4 2015 and Q4 2025. The convergence comes from the same managers, not from a change in the group's makeup: among the 187 managers in the group on both dates, the share holding at least 3 of the top 10 rose from 29.9% to 42.8%. The share holding none held steady (33.7% and 34.2%). The panel of 168 managers present in at least 40 of the 46 quarters (dotted line) shows the same pattern.

The most striking illustration is the trio of Alphabet, Microsoft and Amazon. We chose it after the fact, because these are the three stocks that contribute most to overlap in 2025, so it exaggerates the trend. The share of concentrated managers reporting all three went from 4.9% (21 of 429) to 27.4% (194 of 709), and among the same 187 managers, from 3.2% to 27.8%. Holding all three does not imply large positions: one manager verified on EDGAR holds just 3,000 shares of Amazon (0.07% of its portfolio).

What this means for anyone following 13F portfolios. Before reading a 13F, it helps to place the manager in one of the two groups. Four in ten concentrated managers hold no megacaps at all, so their portfolios are "different from the index" in a literal sense.

2. Outside the megacaps, similarity barely changed

Between 2015 and 2025, average overlap nearly doubled (from 2.15% to 3.97%), and manager pairs sharing at least 20% of their portfolios went from 0.95% to 5.33% of all pairs. Without the top 10, overlap rises only 0.3 points (from 1.55% to 1.83%), and pairs sharing at least 20% go from 0.51% to 0.87%. In Q4 2025, the 10 megacaps account for 65% of average overlap (36% in 2015), and Alphabet, Microsoft and Amazon alone for 49%. With the same 187 managers, the result does not change: 2.46% → 4.33% with all stocks, and 1.73% → 1.90% excluding megacaps.

How to read an overlap between two managers. In Q4 2025, Kensico Capital Management Corp and Gladstone Capital Management LLP share 10.3% of their portfolios, all of it in megacaps: outside of them, they have no stock in common. By contrast, Central Securities Corp and Wedgewood Partners Inc share 28.5%, and even without the top 10 they overlap on 11.5% of what remains. Only the second pair tells you anything beyond "they own the largest stocks".

What this means for anyone following 13F portfolios. When two well-known managers "overlap", subtract the megacaps before reading it as shared conviction. What remains is the informative part.

3. The cost of copying several managers

If a reader combines, in equal parts, the 13F portfolios of 10 randomly chosen concentrated managers, in Q4 2025 the result is a portfolio equivalent to about 76 equal-weighted stocks, versus about 106 if the ten shared nothing (medians of each measure). The median loss, calculated draw by draw, is 25.6% of the apparent diversification (15.3% in 2015). With 20 managers, the loss is 43.0% (28.7% in 2015). Depending on which ten are chosen, the loss ranges from 10% to 45% (10th and 90th percentiles). Almost all of the increase comes from megacaps: without them, the loss with 10 managers goes from 11.4% to 12.8%.

What this means for anyone following 13F portfolios. Adding managers does not add diversification in the same proportion. Beyond about 10 managers, much of what gets added is the same megacaps, which are also the highest-value stocks in the entire 13F universe.

4. They converge on megacaps, but their weight is half that of other filers

Concentrated: 709 managers. Others: 4,644 filers with at least $250 million in stocks that are not concentrated. Average weight with equal weight per manager, on the 13F stock portfolio excluding funds.

The concentrated group is converging on Alphabet, Microsoft and Amazon, but remains far from Apple and NVIDIA. Those two account for 51% of the total megacap weight gap. That is why the group appears underweight in Technology: −8.9 points with equal weight per manager in December 2025. The gap lies mainly in those specific megacaps, though not only there: without the top 10, the mean nearly evens out, but the median remains lower.

What this means for anyone following 13F portfolios. The group's overlap on some megacaps does not bring it closer to other filers. Its megacap weight is half that of other filers, and fewer than a third of the managers report Apple (the same goes for NVIDIA).

Secondary: does the company of other managers protect?

Is a stock held by many concentrated managers safer? We follow the design of GestorLens's own study, recalculated from the positions: 147,065 observations (stock-quarter), 11,100 stocks, 42 quarters from 2015 to 2025. We compare stocks held by 4 to 10 concentrated managers with those held by 1, holding popularity among all 13F filers constant (deciles by quarter) and holding prior volatility constant (terciles).

With those controls, stocks held by 4 to 10 concentrated managers did not have fewer severe declines over the following 12 months. They had a 1.9-point higher probability of falling by half (90% interval, with 4-quarter blocks: 0.8 to 3.2) and also a 2.1-point higher probability of beating the S&P 500 ETF SPY by more than 50 points (1.0 to 3.1). The tails are wider in both directions. Without the volatility control, the difference in falling by half is 3.0 points, and in non-overlapping 12-month windows it is positive in all four subsamples (2.7 to 3.7 points). Part of the effect is explained by these managers already choosing more volatile stocks: the prior volatility of those stocks was between 14% and 70% higher within the same decile. Returns use the price implied by 13F filings, price only, without dividends, capped at +300%. If a stock stops trading, its last price is used, which understates the decline.

What this means for anyone following 13F portfolios. The "company" of other concentrated managers did not work as a safe haven: a stock's risk cannot be inferred from how many managers hold it.

Conclusion

What we observed: between 2015 and 2025, part of the concentrated managers came to report the same megacaps, especially Alphabet, Microsoft and Amazon, and that is why their portfolios look more alike. Another part, close to four in ten, holds none of them. Outside those ten stocks, similarity rose little. The group's weight in them is half that of other filers.

A possible interpretation, not proven: that those stocks grew until they were hard to ignore for managers with similar quality criteria. A 13F does not show the reasons.

Methodology

• Concentrated managers: filers that meet GestorLens's criteria in a given quarter (between 5 and 100 stock positions, top 10 positions making up at least 45% of the portfolio, at least $250 million, average turnover of up to 50%, options and ETFs each up to 20%, and at least 8 quarters of history). Membership is assessed quarter by quarter. There are 1,634 distinct managers across 46 quarters (methodology).
• Reported 13F portfolio: stocks (EQUITY class) with shares and value greater than zero, excluding ETFs and funds, grouped by issuer (Alphabet's Class A and Class C shares count together, and Google's former CUSIP is mapped to Alphabet). Weights are relative to that portfolio, not to the manager's total assets.
• Megacaps: the 10 stocks with the highest value reported by all 13F filers in that quarter, with a filter for scaling errors. The set is fixed before looking at the results.
• Overlap between two managers: the sum, stock by stock, of the smaller of their two weights. It is 0% if they share nothing and 100% if their portfolios are identical. Excluding megacaps means those 10 stocks are removed and weights are recalculated on what remains of each portfolio. That is why the 1.83% "excluding megacaps" is not 3.97% minus the megacap contribution (65%): they are two different measures.
• Calculation: overlap is the exact average over all pairs, with the exact contribution of each issuer. Cost of copying: loss = 1 − effective N of the combination / effective N if they shared nothing, with effective N = 1/HHI, over 2,000 draws per year and number of managers, with a fixed seed. The audit obtained 15.2% → 25.6% and 11.5% → 12.7% with a different random sequence.
• Audit controls: strict panel (the same 187 managers), a version without positions that came in through later amendments (trio at 27.2% instead of 27.4%), excluding closed-end funds (no change), requiring a minimum weight (each of the trio at least 1%: 1.6% → 19.0%), and a manager bootstrap (trio difference +22.5 points, 90% interval: 19.2 to 25.8).
• EDGAR verification: the audit checked 72 share figures from 24 portfolios (4 cited cases and 20 randomly selected managers from 2015 and 2025). All 72 match.
• Correction: in Q2 2021, one filer (CIK 1732080) used nonexistent CUSIPs with mis-scaled values, which erroneously placed the issuer sharing that CUSIP prefix in the top 3. To rank the megacaps, only positions with a consensus price (at least 3 filers) are counted. This changes only that quarter. It does not change any figure for 2015, 2025 or 2026.

Limitations

• A 13F shows a snapshot at quarter-end that becomes public up to 45 days later. Holdings are not current purchases. A 13F does not include short positions, cash, stocks not listed in the US, or other assets.
• Weights are relative to the reported 13F portfolio, not the manager's total assets. Part of the rise in megacap weight is a price effect, because they rose more than other stocks. That is why the headline figures are based on presence, which does not depend on price.
• "Concentrated manager" is GestorLens's own versioned definition. Filer type is a current attribute applied to the entire history.
• Combining managers at random does not represent how a real reader chooses. If the best-known managers overlapped more, the loss would be greater, but that was not measured.
• No figure in this report evaluates average returns or is meant for choosing stocks.

Sources

• SEC Form 13F filings (Form 13F Data Sets and EDGAR XML documents), processed by GestorLens. Data as of Oct 17, 2026.
• Central Securities Corp (CIK 18748), Q4 2025, accession 0001209992-26-000018: Alphabet 420,000, Microsoft 80,000 and Amazon 260,000 shares.
• Kensico Capital Management Corp (CIK 1113000), Q4 2025, accession 0001193125-26-054531: 1,397,600, 338,600 and 1,028,000. Q4 2015, accession 0000950123-16-014949: of the three, only Microsoft (7,439,900 shares).
• Wedgewood Partners Inc (CIK 859804), Q4 2025, accession 0000859804-26-000002: 163,913, 77,496 and 58,266.
• Gladstone Capital Management LLP (CIK 1353394), Q4 2025, accession 0001353394-26-000002: 595,957, 238,707 and 3,000.
• Secondary section: Spirit Realty Capital (CUSIP 84860W102), Q3 2015, reported by 4 concentrated managers in filings 0001140361-15-042438, 0001085601-15-000005, 0001172661-15-001990 and 0001140361-15-041379. The value per share matches the implied price.
• GestorLens's own study: Does copying superinvestors work?, GestorLens, DOI 10.5281/zenodo.23324670.
• On GestorLens: stocks where major investors overlap, methodology and glossary.