調査手法

Universe. The database aims to cover every exchange-traded product listed on major U.S. exchanges. The vendor's ETF list is synced daily; funds that stop quoting on a U.S. exchange, or whose last vendor quote is more than 30 days old (closed, matured, merged or renamed funds), are marked inactive and kept (never deleted), with their last quote date shown. Newly listed funds join automatically.

Quotes. Vendor quotes, refreshed hourly for the whole universe; during market hours a displayed price can be up to about 1–2 hours old. Each quote shows its own timestamp.

Price cross-check. Daily closes from Financial Modeling Prep are compared weekly against Tiingo's for every active fund; a close more than 1% apart is flagged for review. Where FMP's history was confirmed wrong, the affected bars were replaced with Tiingo's, and the fund page notes it under the price chart.

Returns. Computed by ETF Explorer from daily closing data. Price return uses split-adjusted (not dividend-adjusted) closes: close(end) ÷ close(start) − 1. Total return uses the dividend-adjusted close series (distributions treated as reinvested). The vendor re-bases these series whenever a fund goes ex-dividend or splits, so each daily update checks for that and re-loads the fund's full history when it happens. Periods of three years and longer are always annualized: (1+r)^(1/years) − 1. YTD measures from the last close of the prior year. A figure is omitted (—) when the fund's history doesn't cover at least ~90% of the period, or when a data gap of more than three weeks sits at the period boundary. "Since inception" measures from the start of our stored daily series; for a small number of funds this can begin later than the fund's true inception date (the inception date shown on the page is the issuer's). Since-inception returns are annualized when the stored history spans more than about 1.5 years; shorter spans are shown cumulative. Returns above +2,000% for a period are not shown, because moves that large are almost always data problems or extraordinary events.

Trailing yield & distribution growth. Yield (TTM) = cash distributions over the last 365 days ÷ current price. It is NOT the SEC 30-day yield (not supplied by our data source; the issuer's page has it). Distribution growth compares the trailing-12-month sum of regular distributions (special and irregular payouts, such as capital-gain distributions, are excluded) with the prior 12 months for 1-year growth, or with the 12 months ending 3 or 5 years earlier for 3-/5-year growth (annualized). A growth figure is computed only when both windows have a comparable number of payments and the earlier window's distributions are at least 0.25% of the current price; otherwise it is omitted and the fund is excluded from those rankings rather than estimated. Because year-end capital-gain payouts are often labelled as regular in the source data, a single payout more than three times the fund's median regular payout is also treated as a one-off, and annual growth above +500% is not shown.

Risk statistics. From daily returns of the dividend-adjusted series: volatility = standard deviation of daily returns × √252. Sharpe = (annualized total return − 3-month T-bill rate) ÷ annualized volatility, where annualized total return is the compound (geometric) annual rate over the window: (product of (1 + daily return))^(252 ÷ number of days) − 1; the T-bill rate is FRED DTB3 (value shown on each fund page). Sortino (1-year only) uses the same numerator ÷ downside deviation, where downside deviation = √(Σ min(r, 0)² ÷ N, over all N daily returns) × √252 (0% target). Max drawdown = worst peak-to-trough of the adjusted series in the window. Beta (3 years) and correlation (1 year) are computed against SPY as the S&P 500 proxy and need at least 250 and 100 paired daily returns respectively; returns need about 90% coverage of the period, otherwise the statistic is omitted. Volatility, Sharpe, Sortino and drawdown are not shown when the fund's price moved more than 20× within the window (the same cap that hides returns above +2,000%), because such moves are usually data problems or extraordinary events that make the ratios meaningless.

Fund-flow estimates. ETF Explorer snapshots each fund's reported AUM daily. Estimated net flow over a window = AUM(end) − AUM(start) × (1 + price return). This removes the market-move effect but remains an estimate, not official flow data, and it appears with that label everywhere. An estimate is computed only when both AUM snapshots are genuine and recent: not when the vendor's AUM is unchanged between the two dates, looks rounded, or jumps out of line with neighbouring days. Because price return excludes distributions, a fund's own payouts lower its AUM and price together and are not counted as outflows. Issuer-reported AUM that updates only monthly (Vanguard) supports 1-month estimates only.

Trending score. A transparent popularity score, recomputed daily: 2×min(volume ÷ average volume, 6) + 0.8×min(|day move %|, 10) + 0.5×min(news articles tagged to the fund in 7 days, 12) + 0.6×min(page views on this site in 7 days, 20) + 0.6×min(|estimated 1-week flow % of AUM|, 10). Funds under $50M are not scored. Trending measures attention on this site, never investment quality.

Classification. Categories, leverage/inverse flags, currency hedging, option-income and buffer tags, and geography come from deterministic rules over the fund's name and the actual reported sector/country weightings. Sector, bond-type and commodity tags need the fund name or, for unleveraged equity funds, a sector of at least 55% of reported weights; issuer descriptions are used only for option-income and buffer strategies and for active management, because their boilerplate ("financial performance", "may invest in Treasuries") would otherwise mis-tag funds (e.g. "single-country" requires the country to dominate reported weights AND appear in the fund name). Leverage and inverse flags, and the stated multiple (e.g. 3x, −2x), come from the fund name, because issuer descriptions of short-term bond funds and of bull/bear fund pairs would otherwise trigger them; "short-term", "short duration" and "ultra-short" bond funds are not inverse. A fund is also flagged leveraged when its description states a return multiple above 1 (for example "1.2 times (or 120%) the total return of" a stock, as in weekly-pay funds), even if its name does not say so. Leveraged and inverse funds are excluded from the yield, risk and dividend-growth rankings. Active vs. passive is taken from the issuer description where stated; otherwise it is shown as "not stated", never guessed. The underlying index is parsed from the fund name only when unambiguous.

Premium/discount. (Delayed price − last reported NAV) ÷ NAV, labeled indicative because the two figures have different timestamps. The issuer's official premium/discount history governs.

Missing data. "—" or "N/A" means the source has not published a value or our coverage has not reached it yet. We never fill gaps with invented numbers, and AI is never used to generate market data (it only summarizes figures already in the database, with labels).

Refresh cadence. Quotes & news hourly · universe, profiles/AUM, EOD history, top-1500 holdings & distributions, and all derived analytics daily · full-universe holdings and distributions weekly · data briefs after fresh data lands. Each dataset renders its own as-of date.