How StockDrifts rates 13F funds

Every quarter, each 13F filer discloses the US-listed long positions it held at quarter end. StockDrifts takes the positions a fund disclosed at the start of a quarter, holds them for the quarter, and records the return of that basket. Two versions are kept: manager-weighted, where each position counts in proportion to its disclosed value, and equal-weighted, where every position counts the same. Chained over twelve quarters, the manager-weighted version is the fund's three-year return, and it is the single number the rating is built from.

The rating is a percentile of that three-year return among rated filers. Five stars is roughly the top 4%, four and a half stars the next band, and so on down to one star. Because it is a percentile, the rating says where a fund's disclosed book ranks against other filers over the same three years; it does not say whether the fund beat the market. In the three years to June 30, 2026, the S&P 500 returned +68.5%, and every fund rated three and a half stars or better beat that, while no fund rated two and a half or below did. In a different three years the line would sit at a different star.

Why 12 quarters. A fund that has filed for only two quarters still receives a three-year figure upstream, annualized from the quarters it has. A strong two-quarter run therefore extrapolates into a five-star score. Every StockDrifts study restricts itself to funds with at least 12 quarters of filings so that the three-year return is observed rather than projected. In the June 30, 2026 snapshot, 6,522 of 8,725 filers carry a rating, and 5,865 of those have the 12 quarters.

What the rating is not. It is not the fund's NAV return. It is before fees. It ignores short positions, cash, options, non-US listings and every trade made between quarter ends. A long/short fund that hedges its book, or a fund that holds most of its assets outside US equities, can carry a rating that says little about what its investors earned. The rating measures the disclosed stock picks, held a quarter at a time.

The studies published on this site start from the 8,725 filers for the latest quarter and apply four filters. A rating, which requires enough history to compute a return. At least 12 quarters of filings, for the reason above. At most 1,000 positions, because a filer with several thousand positions is an index or a custodian, not a picker. And no match on a list of non-pickers: index managers, custodian banks, market makers and volatility desks, multi-manager quant platforms, sovereign funds, pensions, insurers and corporate holders, whose 13F reflects index rebalancing, hedging or a single strategic stake. That leaves 5,251 active pickers for the June 30, 2026 quarter: 767 rated four stars or better, 996 rated three or three and a half, and 3,488 rated one to two and a half.

When an article names funds, it applies two more filters so that the list is investable: at least $250m of disclosed positions, and eight or more positions, with venture, charitable, endowment and pension filers removed. Aggregate statistics use the full universe.

"Top-rated" or "elite" in our articles means four stars or better with 12 quarters. "Bottom-rated" or "the crowd" means two and a half stars or below with the same filters. The two groups are compared directly because they are the same kind of filer, selected the same way, differing only in their three-year record.

For each fund the database keeps its top 50 positions at the latest quarter, each with the shares held, the value, the weight in the book, a six-quarter history of share counts, and an estimated average purchase price. That estimate is reconstructed from the sequence of disclosed share changes: shares added in a quarter are assumed bought at that quarter's price, shares held before the history begins at their price then. It is an estimate of cost, useful for asking whether a position sits above or below what the fund paid, and it will be wrong for positions built over many years or traded heavily inside a quarter.

Position-level studies remove non-equity rows (ETFs, trust units, notes, warrants, preferreds), and remove rows whose implied price (value divided by shares) differs by more than 2.5 times from the price implied by other holders of the same stock, which catches convertible notes filed under a stock's ticker and positions reported in the wrong units. Industry and sector labels come from Financial Modeling Prep's taxonomy, with StockDrifts' own ticker table as the fallback for the roughly half a percent of tickers FMP lacks.

Because the database keeps the top 50 positions, a fund with 300 holdings is represented by the 50 largest. For four- and five-star funds those 50 positions cover a median 82% of the disclosed book; for bottom-rated funds, which tend to hold more names, a median 38%. Studies of where money sits are studies of where the largest positions sit.

Long only. Form 13F reports long positions in US-listed securities above the reporting thresholds. Short positions are not reported. A fund that is long Micron and short a basket of chip equipment names shows up as a Micron bull.

No cash, no non-US. Cash, bonds, most derivatives and securities listed outside the US are absent. A global fund's 13F may be a fraction of its assets.

Stale by up to 45 days. Filings are due 45 days after quarter end, and the largest wave arrives on the deadline. Positions can have changed before they are public. Every StockDrifts article states the snapshot date and the range of filing dates it used.

Survivorship. A fund that stopped filing before the latest quarter is not in the universe. Scorecards built on current filers therefore omit the funds that closed, which are unlikely to have been the winners.

Amendments and duplicates. Amended filings replace the original. Managers that file under several entities (Capital Group files as three) appear as separate funds; articles say so when one group dominates a result.

Each article in the Q2 2026 series opens with a three-line box summarizing this page and links back here. The same universe, the same rating and the same position filters run through all of them, so numbers can be compared across articles.