Where the data comes from
| Source | Used for | Convention |
|---|---|---|
| Issuer fact sheets and prospectuses | Holdings, weights, expense ratios, index tracked, distribution policy | Primary source, linked directly in the article rather than summarised second-hand. |
| Yahoo Finance price history | Daily total-return series used for every return, drawdown and volatility figure | Adjusted close, so distributions are treated as reinvested on the ex-date. |
| FRED (Federal Reserve Bank of St. Louis) | Interest rates, inflation, spreads and other macro series | Series identifiers are named in the article where a specific series is used. |
| Index provider methodology documents | How an index selects, weights and reconstitutes | Consulted when a fund's behaviour is a property of its index rather than its manager. |
| Kenneth R. French Data Library | Factor return series for market, size, value, profitability, investment and momentum | Used for factor attribution; the regression window is stated where one is run. |
How the figures are computed
| Item | Convention |
|---|---|
| Returns | Total return unless a chart is explicitly labelled price-only. Distributions are assumed reinvested at the closing price with no tax or friction deducted — a simplification that flatters distribution-heavy funds relative to a taxable account. |
| Comparison charts | All series are indexed to 100 at the start of the window so funds with different share prices can be read on one axis. A second y-axis is never used. |
| Drawdown | Decline from the running peak of the total-return series, not from a fixed start. The maximum drawdown quoted is the deepest point inside the stated window only. |
| Window | Five years ending at the article's publication date, unless the article states otherwise. Changing the window changes drawdown and Sharpe figures materially, which is why the window is always named. |
| Annualising | Daily volatility is scaled by √252. This assumes independent returns and therefore understates the severity of clustered, fat-tailed moves. |
| Currency | US dollars, unweighted for an investor's home currency unless the article is specifically about currency exposure. |
| What counts as a holding (definition since 2026-09-18) | Every weight statistic — effective holdings, top-ten share, largest position, overlap and the calculator — is computed on the fund's own investments: the filed positions with securities-lending cash collateral set aside, renormalised to their own total. A fund that lends stock receives cash and parks it in a money-market fund; the filing lists that as a position (flagged isCashCollateral) but it is matched by a liability and is not something the fund chose to own. Counted as a holding it dominated the arithmetic for 211 of the funds measured here, in every quarter back to 2019. IJH, for example, would read 202.1 with its 3.84% collateral line counted; on its own investments it reads 287.4. Each record keeps the as-filed figures under asFiled, so the numbers published before this date remain reproducible, and 331 funds carry a note stating the collateral share and the as-filed reading. Between 2026-09-14 and 2026-09-18 the headline still counted the collateral and 73 funds were marked not comparable instead; that interim rule is retired. |
| Filing weights | Positions are weighted by the pctVal each fund files. Negative positions are dropped, so a fund holding offsetting shorts reports weights summing above 100% — that is the concentration of what it holds long, which is the quantity being measured. Shares within a fund are normalised to its own filed total, because those totals rarely land on exactly 100 and an unnormalised share cannot be compared between two funds. |
| Comparability | A fund's concentration figure is treated as comparable only when its filed long positions cover at least 95% of net assets and derivative exposure is under 20%. Outside those bounds the filing does not describe the portfolio the way it does for an ordinary fund: one managed-futures fund filed twelve derivative positions summing to 1.42% of net assets, and normalising that fraction produced an effective holding count near 37,000. Both thresholds were set by scanning every fund-quarter in the archive, not chosen by eye, and no ordinary fund is caught by them. Securities-lending collateral is handled by the definition above rather than by exclusion. A fourth rule, added 2026-09-19: when swap and futures notional reaches 50% of net assets the fund is not comparable, whatever its filed derivative value — a 2x fund files its swaps at their unrealised gain (SSO: 13% of value against 134% notional), so the earlier value-based test let 49 geared funds through with a position count that does not describe their exposure. Excluded funds keep their published row and their numbers with the reason attached; they are held out of headline figures and rankings only. Currently excluded: TQQQ, SOXL, NVDL, BBJP, BOND, QLD, BOXX, CGCP, SSO, SPXL, EVTR, AMDL, PDBC, MULL, UPRO, SQQQ, BITX, MVLL, DBMF, MAGS, HGER, ETHU, INTW, TSLL, USD, FTGC, FAS, BCI, TMF, TSLQ, LDUR, VPLS, CONL, BITO, TNA, WCMI, ALLW, CTA, FJAN, JFLX, FDEC, XAGG, KORU, FFEB, HFSI, ROM, NTSX, FSEP, FNOV, FMAY, FOCT, FJUL, FAUG, LITX, NBSD, FAPR, SH, COMT, HYGH, BUFC, NEBX, MSOS, MSFU, CERY, PTRB, ACYN, DLLL, MSTY, NUGT, UDOW, PSQ, NOWL, CMDT, YINN, APPX, TUA, UYG, ASTX, SPXU, LABU, FEPI, QMAR, SDS, AGOX, CRDU, HCMT, FBL, CMDY, NAIL, DPST, NVDU, CSM, SDCI, DDM, BTAL, IONL, DJAN, CWVX, OCTW. |
| Cash-like and unclassified | A fund holding 50% or more in short-term investment vehicles is flagged rather than excluded — its holding count measures something different from an equity fund's. The same applies when 5% or more of net assets carries no asset category in the filing. |
Definitions
Each measure used across the site has a single definition page carrying its formula, how to read a value, and what it cannot tell you. There are currently 25 of them. Browse the measures →
Data access
Every dataset on this site is downloadable, not just readable — a table on a page is the same numbers a script can pull directly, in JSON or CSV, licensed CC BY 4.0. No key, no rate limit, no signup.
| Dataset | JSON | CSV |
|---|---|---|
| Concentration (effective holdings) | concentration.json | concentration.csv |
| Concentration over time (quarterly) | history.json | history.csv |
| Valuation (fair value levels) | valuation.json | valuation.csv |
| Pairwise overlap | overlap.json | overlap.csv |
| Reverse fund ownership (which funds hold which companies) | ownership.json | ownership.csv |
These are static files regenerated with every update of the site, not a live API — there is no versioning or uptime commitment beyond the site's own. If a stable, documented endpoint would be useful for something you are building, the contact page is the way to ask for one.
What these figures cannot support
- Every figure is as of the publication date stamped on the article. Markets move; nothing here is re-computed on a schedule, and an old article is a record of what was true then.
- Comparisons are drawn from funds that exist today. Funds that closed or merged are absent, which flatters the surviving set — the standard survivorship problem, present here as everywhere.
- This site describes what has happened. It does not forecast returns, and where an article discusses expected values it is describing a model's output, not a prediction.
- Nothing here accounts for an individual's tax position, time horizon, or existing holdings, and none of it is personalised advice.
Corrections
Errors are corrected in place and the article's revision date is updated. If a correction changes a conclusion rather than a detail, the change is stated in the article itself rather than made silently. Reports of data errors, broken sources or disputed figures are welcome via the contact page.
How this is built
This site's data pipelines, tools and much of its prose are built with AI assistance (Claude). The figures are not: every number traces to a cited source — SEC N-PORT filings, Yahoo Finance, FRED — and can be independently recomputed from it. AI writes the pipeline; it does not invent the data, and a tool's stated limits (see "Sources and limits" on each tool page) are verified against the underlying figures before publication rather than asserted from confidence alone.
Independence
This site carries no affiliate links, no sponsored placements and no paid fund coverage. Where display advertising runs, it is served by an ad network with no influence on which funds are analysed or what the analysis concludes. No issuer reviews an article before publication.