Does a top 13F track record persist? About a third of it does

Rank 5,004 funds by the return of their disclosed book from July 2023 to December 2024, cut them into fifths, and follow each fifth through January 2025 to June 2026. If records carried no information, every cell of the transition matrix would read 20%. The top quintile's row reads 32% stayed top, 21% fell to the bottom, with the three middle quintiles sharing the rest. The bottom quintile's row is its mirror: 29% stayed bottom, 23% jumped to the top.

That U-shape at the corners is the main feature of the data. The funds at either end of the first half are disproportionately the funds at either end of the second half, in both directions. What persists most reliably is not the sign of a fund's outperformance but the size of its swings: concentrated, high-variance books stay high-variance.

Run the same test on equal-weighted books, which isolates stock picking from position sizing, and the correlation rises slightly to +0.13, with 34% of top-quintile pickers staying top. Picking persists a little more than sizing does.

The median second-half return rises with the first-half quintile, from +20.9% for the bottom fifth to +25.7% for the top fifth, and the share beating the index rises from 31% to 45%. So the record ranked the funds in the right order. It did not rank them above the index: the S&P 500 returned +27.5% in the second half, and the median fund in every prior quintile, including the top one, came in below it.

Averages tell a slightly different story. The top quintile's mean second-half return was +33.4%, against a median of +25.7%: a minority of repeat winners with very large second-half returns pulls the mean above the index while the typical repeat fund sits below it.

Replace quintiles with a simpler test: did the fund beat the S&P 500 in the prior window, and did it beat it in the next? Half to half, 43% of prior beaters beat the index again and 29% of prior laggards did, against 34% of all funds. Year to year the gap is the same shape: 47% against 23% from the first year to the second, 41% against 32% from the second to the third. Ranking funds on a full three-year record to December 2024 gives 44% against 31% over the following 18 months.

The gap is consistent and it is about 14 points. A prior beater is roughly one and a half times as likely to beat the index next period as a prior laggard, and less likely than not to do it.

Pool the two consecutive one-year transitions (10,135 fund-years) and the rank correlation is -0.04: nothing, on the face of it. But the corners of the matrix are as far from 20% as they are in the half-to-half version. Top-quintile years were followed by another top-quintile year 34% of the time and by a bottom-quintile year 25% of the time; bottom-quintile years were followed by another bottom year 33% of the time and by a top year 27%.

A near-zero correlation with strong corners means the same thing as before. Ranks do not persist in a straight line; what persists is membership of the extremes. Over one-year windows, the funds that produce top-quintile years are largely the funds that also produce bottom-quintile years.

If each year's result were an independent draw at that year's beat rate (31%, 30%, 35%), 3.2% of funds would beat the index in all three years. 6.2% did, 311 funds. Being in the top quintile in all three years would happen 0.8% of the time by chance; 126 funds (2.5%) did it. Being in the bottom quintile in all three years: 145 funds (2.9%).

The five-year version, for the 3,769 funds with twenty quarters of history, is the cleaner test because the window includes the 2022 bear market. 21 funds beat the index in all five years against 11 expected; 194 did it in four of five against 127. And at the other end, 757 funds beat it in none of the five years against 560 expected. Both tails are fatter than independence allows. Skill and its absence both show up in the streaks, and the absence shows up more.

StockDrifts' star rating is a percentile of a three-year record. To test what such a rating would have told an investor, rank the 4,448 funds with enough history on their three-year return to December 2024, the window a rating computed then would have used, and follow them through June 2026. The top quintile of that prior record beat the index over the next 18 months 46% of the time, the bottom quintile 34%, all funds 34%. Narrow it to the top 10% and the beat rate is 51% with a median return of +27.7%; narrow it to the top 4%, the five-star equivalent, and it is 54% with a median of +30.6%, against the index's +27.5%.

So a five-star record was worth something: a better-than-even chance of beating the index over the following 18 months, in a period when two thirds of funds did not. It was not worth a lot. The rank correlation between the prior record and the follow-up is +0.10, and only the top decile of the prior ranking beat the index at the median.

240 funds with at least $250m disclosed and eight or more positions were in the top quintile in both halves, out of 1,001 top-quintile starters. The twenty-five with the strongest second half are below, each linked to its StockDrifts page. Several are Asia-based managers of US technology supply-chain books (FengHe, WT Asset Management, CloudAlpha, Hel Ved); several are semiconductor specialists (Analog Century, Atreides); and two are large multi-strategy books whose disclosed longs happened to run (Whitebox, Castleknight).

Three years in a row is the stricter test. 95 investable funds were in the top quintile in each of the three years to June 2026; the thirty with the strongest third year are in the second table.

117 investable funds went from the top fifth in the first half to the bottom fifth in the second. The largest are household names: Baillie Gifford ($110bn disclosed, +50% then +3%), Loomis Sayles, GQG Partners (+81% then +6%), Horizon Kinetics (+121% then +12%). The reverse move, bottom fifth to top fifth, was made by 142 investable funds.

Two cautions on reading the names. These are returns of the disclosed US long book, not fund returns; a global or hedged manager's 13F is a slice of its portfolio. And a large fund whose book returned +3% in eighteen months when the index returned +28% has had a poor stretch by this measure, not a disaster. The point of the list is that the first-half leaders included some of the largest, longest-established growth managers in the database, and the second half reversed them.

Search all 5,004 funds. Quintile 5 is the top fifth of the 5,004 in that half, quintile 1 the bottom fifth. The three yearly columns are the three consecutive years to June 30, 2026.