Methodology · Study
Does copying superinvestors work? 129,534 cases since 2013
By Cristian Díaz · GestorLens · · revised on October 12, 2026 · version 1.4.0 of the calculation
We took every stock held by concentrated managers each quarter since 2013 (129,534 cases) and looked at the next 12 months against the S&P 500. Short answer: on average, they did not beat the index; a few specific groups show a gap in their favour, small and seldom firm.
Only 40% of cases beat the S&P 500 over 12 months (129,534 cases, 2013–2025).
What was measured, and what was not. We measured the simple average of the stocks held by any concentrated manager: every stock weighs the same, regardless of how much it weighed in each portfolio or who held it. We did not measure copying an investor's portfolio with its weights: that is measured, manager by manager, in the ranking. Nor is it a study of “the best investors”: concentrated managers are defined by the shape of their portfolio, not by fame or results.
Limits of this study
- Price only. No dividends, for the stocks or for the index. If these stocks pay less than the S&P 500, the result with dividends would be somewhat worse; we did not measure it.
- Prices from the 13F filings themselves. Reported value divided by shares, one per quarter; they are not market quotes. Splits are detected in the filings (9 of the 19 corrected ones were checked against quotes and matched).
- Stocks that stopped trading (6.0% of cases): counted at their last price, which in a bankruptcy overstates what holders received.
- No costs, and the purchase is assumed at the end of the following quarter, not on the day of publication.
- The group is defined by the shape of the portfolio. It is not a selection of the best.
- 52 groups tested on the same sample: a t close to 2 is a hint. The headline figure (t = −1.56) is not significant.
- One period, 2013–2025, with an index dominated by a few giant companies, compared with an equal-weighted average.
129,534
cases: one stock in one quarter
8,228
distinct stocks
48
quarters (2013–2025)
40%
of cases beat the S&P 500
What we found
- Averaging the quarters (every quarter weighs the same), the stock held by concentrated managers returned 3.7 points less than the S&P 500 over the following twelve months, and only 40% of cases beat it. It is a simple average (every stock weighs the same) against a size-weighted index dominated by a few giant companies that few concentrated managers hold with weight.
- The link with accumulation grows with time, but the 4+ quarters group does not reach the usual threshold: against the universe, stocks that managers had been adding to returned +0.5 points a year with one quarter of accumulation, +1.5 with two, +1.7 with three and +2.2 with four or more. The highest t is the two-quarter one (t = 4.65). Four or more has the largest gap, but it stands at t = 1.88 and beat the universe in 61% of quarters: it is below the usual threshold (|t| ≥ 2) and we do not present it as a proven effect. Stocks nobody was accumulating lagged consistently (t = −3.60; they beat the universe in 29% of quarters). Caveat: we tested 52 groups, and with that many tests even a t above 2 is a hint, not proof.
- Combining “more net managers” with a profitable company gave +3.0 points over the universe and beat it in 74% of quarters, though with too few cases to be certain (t = 1.26). Loss-making companies had a better average (+6.6 points) but a worse median (−14.8% versus −7.9%): a few extreme cases lift the average while most lose. That result should not be read as significant: only 44% of cases have financials available, and they are companies that still exist today.
- “More net managers” alone is not enough: the change in the number of managers holding the stock, on its own, showed no reliable edge. Neither did the margin of safety (buying below an estimated intrinsic value) in our second study.
- The 308 cases where more than 100 concentrated managers held the same stock returned +8.8% a year over the S&P (t = 2.27): these are the decade's big winners, and there are few of them. They are the same giant companies repeated quarter after quarter, with overlapping 12-month windows: the independent cases are far fewer.
The numbers
Universe: all stocks held by concentrated managers. Its 12-month excess versus the S&P 500 is −3.7% (average of quarters): that is the study's headline figure, calculated the same way as each group. That gap versus the index is not statistically significant (t = −1.56). Each row is compared with that universe in its own quarters, not with one fixed figure. Secondary figure, which is not the study's figure: averaging all cases together gives −2.8%. It differs because quarters with more stocks, the most recent ones, then weigh more.
No quarter of accumulation|t| ≥ 2
−0.5 pp versus the universe · t −3.60
Excess over the S&P 500: −4.2% a year (universe in the same quarters: −3.7%). It beat the universe in 29% of quarters; 39% of its cases beat the S&P 500. 80,484 cases.
1 quarter of accumulation
+0.5 pp versus the universe · t 1.36
Excess over the S&P 500: −3.1% a year (universe in the same quarters: −3.7%). It beat the universe in 58% of quarters; 41% of its cases beat the S&P 500. 25,178 cases.
2 quarters in a row of accumulation|t| ≥ 2
+1.5 pp versus the universe · t 4.65
Excess over the S&P 500: −2.1% a year (universe in the same quarters: −3.7%). It beat the universe in 73% of quarters; 42% of its cases beat the S&P 500. 11,424 cases.
3 quarters in a row of accumulation
+1.7 pp versus the universe · t 1.91
Excess over the S&P 500: −1.6% a year (universe in the same quarters: −3.3%). It beat the universe in 65% of quarters; 41% of its cases beat the S&P 500. 5,699 cases.
4+ quarters in a row of accumulation
+2.2 pp versus the universe · t 1.88
Excess over the S&P 500: −1.0% a year (universe in the same quarters: −3.2%). It beat the universe in 61% of quarters; 40% of its cases beat the S&P 500. 6,737 cases.
More net managers (Δ ≥ +3) and profitable company
+3.0 pp versus the universe · t 1.26
Excess over the S&P 500: −0.1% a year (universe in the same quarters: −3.1%). It beat the universe in 74% of quarters; 44% of its cases beat the S&P 500. 1,664 cases.
More net managers (Δ ≥ +3) and high conviction
+2.2 pp versus the universe · t 0.72
Excess over the S&P 500: −0.3% a year (universe in the same quarters: −2.5%). It beat the universe in 60% of quarters; 44% of its cases beat the S&P 500. 2,300 cases.
Only “more net managers” (Δ ≥ +6)
+3.2 pp versus the universe · t 0.78
Excess over the S&P 500: +0.9% a year (universe in the same quarters: −2.3%). It beat the universe in 51% of quarters; 42% of its cases beat the S&P 500. 760 cases.
Held by more than 100 concentrated managers|t| ≥ 2
+11.9 pp versus the universe · t 2.27
Excess over the S&P 500: +8.8% a year (universe in the same quarters: −3.1%). It beat the universe in 77% of quarters; 57% of its cases beat the S&P 500. 308 cases.
Highest conviction (top quintile of median weight)
+1.4 pp versus the universe · t 1.07
Excess over the S&P 500: −1.2% a year (universe in the same quarters: −2.7%). It beat the universe in 67% of quarters; 42% of its cases beat the S&P 500. 12,785 cases.
Company with positive operating income (with financials available)
+3.2 pp versus the universe · t 1.57
Excess over the S&P 500: +0.7% a year (universe in the same quarters: −2.5%). It beat the universe in 60% of quarters; 40% of its cases beat the S&P 500. 38,398 cases.
Company with operating losses (with financials available)
+6.6 pp versus the universe · t 1.61
Excess over the S&P 500: +4.2% a year (universe in the same quarters: −2.4%). It beat the universe in 64% of quarters; 39% of its cases beat the S&P 500. 19,068 cases.
- Cases (N):
- one stock in one quarter.
- Excess over the S&P 500, per year:
- 12-month price return minus the S&P 500's, averaged by quarter and then across quarters.
- Universe in the same quarters:
- the same for the whole universe, in the quarters where the group has cases: that is why it changes from row to row.
- Gap versus the universe:
- the group's excess minus the universe's: what the criterion adds.
- t:
- t-statistic of that gap (Newey-West, 3 lags). |t| ≥ 2 is flagged, and “≈ 2” if it rounds to 2.0; with 52 groups tested, that threshold is not enough to claim an effect. “Biased sample”: it uses financial statements and only includes companies that exist today; its t is not flagged.
- Quarters it beat the universe:
- counted by quarter, not by stock.
- Cases that beat the S&P 500:
- counted by case (stock and quarter). A different measure: the two used to share the “Wins” column.
Why the groups of a criterion do not always add up to the total. A group does not count the quarters in which it had fewer than 5 stocks, some criteria are measured on part of the universe, and the profitability ones have no data without financials. Accumulation: 129,522 in the table + 12 in quarters where the group had fewer than 5 stocks = 129,534. Profitability: 57,466 in the table + 24 in quarters where the group had fewer than 5 stocks + 72,044 with no published financials = 129,534. The study's program does that count for every criterion and stops if any of them does not add up.
To rebuild the table: observations (Parquet) · CSV · column dictionary and how to verify
The working paper (PDF) and these observations are deposited at Zenodo, with a permanent address: DOI 10.5281/zenodo.23324670
The same measurement over other horizons
6 months
−1.5%
versus the S&P 500 (average of quarters); t −1.16 against zero. 42% of cases beat the index.
135,807 cases in 50 quarters (2013–2025)
12 months · the study's figure
−3.7%
versus the S&P 500 (average of quarters); t −1.56 against zero. 40% of cases beat the index.
129,534 cases in 48 quarters (2013–2025)
24 months
−11.0%
versus the S&P 500 (average of quarters); t −4.10 against zero. 34% of cases beat the index.
116,999 cases in 44 quarters (2013–2024)
Each horizon uses every quarter that already has an exit price, so the period is not the same. The 24-month figure is cumulative over two years, not annual, and at that horizon the +300% cap and delisted stocks weigh more. The t is against zero: below 2 in absolute value, the gap with the index does not reach the usual statistical threshold.
How we measured it
- Source: every SEC 13F since 2013. Concentrated managers: 13F filers whose portfolio has 5 to 100 stocks, at least 45% in the top 10 positions and low turnover. They are investment firms, family offices and insurance holding companies, plus filers that meet those criteria and whose type we could not verify with a document. Full definition
- Each criterion uses only the information available the day the 13F was published (up to 45 days after quarter end), with no look-ahead.
- Delisted stocks are included (6.0% of cases): without that adjustment results look better than they are.
- No costs or taxes. Past results do not guarantee future ones. Not investment advice.
Technical notes and limits
- Prices: the price implied by the 13F filings themselves at each quarter end (reported value ÷ shares, with at least three filers), split-adjusted. These are not market quotes: there is one price per quarter, not per day.
- Entry at the end of the quarter after the portfolio date (when the 13F was already public) and exit twelve months later. The criteria are assessed with that window; the universe was also measured over 6 and 24 months.
- Price return, without dividends, for both the stocks and the S&P 500 (measured with the SPY ETF).
- Capped on the upside only: each case's return is limited to +300%, so that a few extreme cases do not move the average. On the downside the natural limit is −100%; the cap is not symmetric and lowers the average of groups with extreme cases.
- Stocks that stopped trading (takeover or bankruptcy) are valued at the last available price. In a bankruptcy that price is usually above what shareholders received: the adjustment only partly corrects the upward bias.
- The 12-month windows overlap (there is one per quarter): cases are not independent, which is why t uses Newey-West errors. 3 lags are used.
- It is a simple average of stocks compared with a capitalisation-weighted index: part of the gap comes from that. The result weighted by portfolio weight was not calculated.
- Stocks that disappear before having any price after entry are left out of the measurement.
- Multiple comparisons: 52 groups were tested. With that many, some exceed |t| = 2 by chance: results with t close to 2 are hints.
- Profitability: 57,490 of the 129,534 cases (44%) have financials available, and they are companies that still exist today. We use the latest fiscal year closed at least 90 days before entry, with figures as they stand today (they may have been restated): an approximation of “what was known at the time”, not the exact data.
- The consensus score weights come from this same sample: there is no test on later data yet. Of its four factors, this study supports only one, accumulation.
- No trading costs or taxes. It describes the past and does not anticipate future results.
Version 1.4.0 note: until October 10, 2026 the site quoted the average of all cases together (−2.8%) as its headline. From this version on there is a single headline, −3.7% over 12 months (average of quarters), and the table states how each criterion adds up. See the correction
How to cite this study
Díaz, C. (2026). Does copying superinvestors work? 129,534 cases since 2013. GestorLens. https://www.gestorlens.com/en/estudios/sirve-copiar-superinversores
Working paper, in English and Spanish, with the data to rebuild the tables: https://doi.org/10.5281/zenodo.23324670
The download files and the study (with its tables, charts and observations) are published under Creative Commons Attribution 4.0 International (CC BY 4.0): you may use, copy, adapt and redistribute them, including commercially, with attribution (“GestorLens (gestorlens.com), with SEC data”). The original SEC documents are in the public domain. The rest of the site, the GestorLens brand and the design are not under that licence. This is not investment advice. CC BY 4.0 · data licence
Use it to research, not to copy
In Ideas, the consensus score ranks what to look at first. It weighs accumulation, which this study supports, and three other data points that it does not: it is explained right there.