WhaleCreep

How WhaleCreep reads 13F filings

Every holding figure on this site is computed from the managers' own Form 13F filings, fetched from SEC EDGAR, and nothing is estimated from anywhere else. The one outside input is a reference price series, used for the stock changes on the discovery section and described below. This page explains what is measured, what is deliberately not, and the three data problems that have to be solved before any 13F figure can be trusted.

What a 13F is

Form 13F is a quarterly report that institutional investment managers with at least $100 million in US equity assets must file with the SEC within 45 days of each calendar quarter-end. It lists long positions in US-listed equities and certain options at their quarter-end value. It does not include short positions, swaps or most derivatives, cash, leverage, or non-US holdings.

What is measured, and what is not

  • Holdings and weights. Each position as a share of the manager's reported equity book. Options are reported at underlying notional and excluded from the equity book by default.
  • Changes in shares, not value. Value moves with price, so a book that did nothing shows bigger numbers after a rally. Comparing share counts isolates the manager's actual decision. A change is only computed against the immediately preceding filing; a gap of more than one quarter voids the comparison.
  • Concentration. Largest position, top-ten weight, and effective positions (1/HHI: the number of equal-weight positions the book behaves like). Reported for the latest quarter and as a median across every quarter filed, so one unusual quarter neither qualifies nor disqualifies a manager. The concentration gate is a median largest position of 5%.
  • Holding tenure. Completed holding spells only. Positions still open have an unknown final length and are reported separately as lower bounds, because pooling them understates exactly the long-held conviction names.
  • Crowding. Names held by many managers at a meaningful weight of their own book, and, more usefully, names several managers opened in the same quarter.
  • Concentration signals. A manager's latest filing against its previous one, when the two are no more than 190 days apart. The book became sharply more concentrated when effective positions fell by 30% or more, or the top-ten weight rose by 10 percentage points or more; less concentrated on a rise of 30% or more, or a fall of 10 points or more. A filing that crosses one way on one measure and the other way on the other is not flagged.
  • Fund vs fund. Two books compared for a quarter both managers filed. Portfolio similarity is the sum, over names both hold, of the smaller of the two weights: the share of capital positioned the same way. Beside it: holdings overlap (shared names over all names, ignoring weight), sector similarity (100% minus half the summed gap between the sector mixes, unclassified names left out) and concentration similarity (how close the two HHIs are). The overall score is their mean. Factor and geographic exposure are not scored, because a 13F carries neither.
  • Not performance. A 13F is long-only US equity with no shorts, derivatives, leverage, cash or international, so a manager's real return cannot be derived from it. WhaleCreep does not publish 13F-derived "returns" and does not rank managers by them.

Three things that are wrong in raw 13F data

1. The value column changed units

Form 13F's value field was denominated in thousands of dollars before the 2023 amendment and in whole dollars after, and filers switched on their own schedules. Applying one rule to the whole history produces a set that is right for some fund-quarters and wrong by exactly 1000× for others. WhaleCreep detects the units per filing by arithmetic: value divided by shares is an implied share price, and its median across a filing must land where real equities trade. The original value and the divisor applied are both kept, so every correction is auditable.

2. Amendments come in two kinds

A 13F-HR/A can either add positions omitted from the original or restate the whole table. Summing every accession double-counts restatements; keeping only the newest discards a full original in favour of a one-row addition. Each amendment is classified by how much its CUSIP list overlaps what is already on file, and one effective book is kept per quarter.

3. Issuer names do not resolve to tickers

13F issuer fields are abbreviated and truncated. Positions are mapped to tickers on CUSIP, the identifier the filing actually gives, via OpenFIGI, restricted to US listings so a foreign line cannot win over the US listing the filing refers to. Sectors then come from the ticker.

Who discovered it first

Each stock page opens with the tracked managers that held a meaningful position early, how those positions evolved, and when ownership broadened. "First reported" is the first quarter-end filing that shows the position; 13F data is quarterly, so no purchase date is claimed. Three rules keep the ranking honest.

  • Observed entries only. A name already in a manager's first filing on record was bought at some unknown earlier date. A manager is ranked only if it was seen filing without the name within two quarters before its first position; everyone else is listed as already holding when coverage began, and never ranked.
  • Meaningful, not merely first. Ranking is by the first quarter the position reached 1% of the manager's book in a run of at least two filings, then by peak weight, so a scrap that vanished after a quarter does not outrank a large, persistent position opened one quarter later. "High conviction" is a peak weight of at least 5%.
  • The run (beta). Where a stock broke out: a quarter-end low it never closed below again, followed by its largest one-year move, which has to be at least a doubling, above the stock's own usual best years (its 90th percentile), and at least three times the starting price at the peak after it. A steady compounder has no run. A manager bought before the run if its position reached 1% of its book in the year before the run began or in its first quarter (a 13F only shows quarter ends), it was seen filing without that weight just before, and it still held at the start.
  • Relative to coverage. The tracked universe widened from about a hundred managers to over a thousand in Q3 2024. Adoption is therefore shown as the share of that quarter's filers holding the name, and the descriptive phases are cut where that share passed one third and two thirds of its peak. They describe the order of adoption, not a recommendation.

Stock changes describe the stock, never a manager's return. They are split-adjusted changes in the quarter-end price, excluding dividends. The filings carry a price (value ÷ shares), which works where a name has many holders and breaks where an early name is held by one or two. A reference series of quarter-end closes is therefore kept per ticker and used only where it agrees, quarter by quarter, with the price the filings imply; a series that disagrees, as with a reused ticker, is set aside and the page says which source it used. Where neither gives a continuous answer the page shows a dash rather than a number.

How the manager list is built

The universe is seeded from public reputation and manager lineage across seven styles: Tiger lineage, activist, concentrated value, quality compounders, macro and discretionary, permanent capital, and emerging concentrated managers. Where a public ranking supports a name, the source and edition are recorded. Long-run outperformance is treated as a claim, never as a measurement; concentration is measured directly from the filings and is what the site's rankings use. A few deliberately diversified quant books are included as a control that the concentration metric works.

Freshness

EDGAR is checked every day. A 13F describes holdings as of quarter-end and is filed up to 45 days later, so the newest filing is between 45 and 135 days old at any moment. Every page states the quarter its figures describe and the date the filing was made.

Caveats, in one place

Pages on this site show figures without repeating these; they apply everywhere.

  • A 13F is a quarter-end snapshot filed up to 45 days later. It lists long US equity positions only: no shorts, swaps, cash or most foreign listings, so it shows neither a fund's current positions nor its performance.
  • Options are reported at underlying notional and are excluded from the equity book, weights and counts unless stated.
  • Position changes are measured in shares against the manager's previous filing, so price moves do not read as decisions. A first filing, or a gap of more than two quarters, has no comparison and no action.
  • A new position in crowding counts only at or above the minimum weight of that manager's own book, so conviction is judged relative to the manager, not in dollars. A fund that did not file the previous quarter cannot have initiated anything. Consensus holdings are mostly index names; simultaneous initiations carry the information.
  • Holders are tracked managers whose filing for the quarter reports the security. The tracked universe widened from about a hundred managers to over a thousand in Q3 2024, so ownership over time is also shown as the share of managers filing that quarter, which is comparable across the change where the raw count is not.
  • Tenure counts completed holding spells only; positions still open have an unknown final length. Effective positions is 1/HHI.
  • Sectors cover positions whose ticker resolved to a sector; the remainder is unclassified, not cash.

Market data

  • Every instrument is shown with the symbol it trades or publishes under: SPY is the S&P 500 ETF, not the index. Levels are the instrument's own.
  • Change is the move over the timeframe: percent for prices and indices, basis points for yields and spreads (4.00% to 4.20% is +20bp). Acceleration is this window's change minus the previous window's; the full history has no previous window and so no acceleration.
  • Each instrument is measured on its own trading days, ending at its own latest close, so the same figure appears on every view. Monthly and weekly series count their own prints: a one-month change on a monthly series is one print.
  • Credit spreads, volatility and financial stress are coloured the other way round: a rise is shown as negative.
  • Prices come from free end-of-day sources (Yahoo Finance, FRED). A series that has stopped updating shows its last level and no change.

Questions

What is a 13F filing?

Form 13F is a quarterly report that institutional investment managers with at least $100 million in US equity assets must file with the SEC within 45 days of each calendar quarter-end. It lists their long positions in US-listed equities and certain options. It does not include short positions, most derivatives, cash, or non-US holdings.

How current is 13F data?

A 13F describes holdings as of the quarter-end and is filed up to 45 days later, so the newest filing is between 45 and 135 days old at any given moment. WhaleCreep checks EDGAR every day and shows the date each filing was made.

Can 13F filings tell you a hedge fund's performance?

No. A 13F is long-only US equity with no shorts, derivatives, leverage, cash or international positions, so a fund's real return cannot be derived from it. WhaleCreep measures concentration and holding tenure, which a 13F does support, and never presents long-book replication as performance.

Why do position changes use share counts rather than dollar values?

Dollar value moves with the share price, so a book that did nothing shows bigger numbers after a rally. Comparing share counts isolates the manager's actual decision to buy, hold, trim or sell.

What does 'effective positions' mean?

It is the reciprocal of the Herfindahl index of portfolio weights: the number of equal-weight positions the book behaves like. A manager with six real bets and ninety tiny positions has ninety-six holdings and an effective count near six.