Data notes
Every table, chart and ranking on this site is produced automatically from price series and public filings. This page explains, for each kind of data, how the figures are calculated, what they refer to and where they stop being reliable. The short version of each note appears as a footnote directly under the block it belongs to.
Two things apply throughout. Analysis and presentation are partly AI-assisted; figures are calculated by code, while wording and graphics may be generated. And none of it is investment advice. Past performance says nothing about future performance.
Where the data comes from, how often it is refreshed and why our figures may differ from other providers is set out on the Methodology and data sources page.
Markets and overviews
Index returns and the 52-week range
Each index is represented by a price series; where no series exists for the index itself, a widely traded ETF on that index stands in for it. Returns are pure price changes — dividends, taxes and trading costs are not included, while for ETF series the fund's own costs are already reflected in the price. The 52-week range shows where the current price sits between the low and the high of the last 252 trading days.
Data as of:
US stock market breadth
Above the 200-day average: latest close above the mean of the last 200 closes. Near the 52-week high or low: within 5%, measured over the last 252 closes. Sector medians use the returns shown on the stock pages and appear once a sector has five stocks. Coverage follows the funds we track and is not the complete US market.
Stock rankings
Gainers and losers come from the 500 largest stocks in the MSCI World plus all German stocks for which we keep a sufficiently long price series. The full table on the stocks overview covers every stock with a price series; neither is the whole market. The return is the pure price change over 12 months, excluding dividends, taxes and trading costs. Corporate actions such as splits or spin-offs can distort individual names.
Data as of:
Crypto prices
Crypto trades around the clock, so the daily close is a fixed cut-off rather than a market close. Prices refer to US dollar trading in the most liquid pairs — other venues and other currencies can differ.
Data as of:
Currency rates
Currency pairs are represented by standardized futures contracts because those have a single, verifiable daily close. Spot rates as quoted by a bank can differ slightly — through the forward premium, trading hours and the spread.
Data as of:
Commodities and precious metals
Commodities and precious metals have no single spot price, so the most heavily traded running futures contract stands in for the market. When that contract expires the next one takes over — which produces a jump in the series that is not a real daily move. We do not smooth these rollovers out; in the daily reports, however, outliers of this kind are kept out of the headlines.
Data as of:
ETF master data: costs and size
Total expense ratio and fund size are master data of each fund, sourced from commercial providers and re-read regularly. Fund size moves with prices and inflows, so it is a snapshot. The selection covers only the European-traded ETFs we track — not a complete market directory.
Data as of:
Interest rates, inflation and the labour market
Interest rate, inflation and labour market series come from the publications of central banks and statistical offices and are neither converted nor smoothed. The as-of date is shown per series because the series are published at different frequencies — daily, monthly or quarterly. Later revisions by the agencies are picked up with the next refresh. The change is the gap to the previous value of the same series and carries its own unit: for rate, inflation and unemployment series that is percentage points (pp), because the gap between two percentages is not itself a percentage — 2.00 to 2.25 per cent is 0.25 percentage points. For the ECB reference rate EUR/USD the change is in dollars, and both level and change are shown to four decimal places, as a daily currency move would otherwise round away.
How many funds hold a stock
The figure comes from the holdings files of the funds we cover: for each reporting date we count how many of them hold the stock and add up their weights. That sum stays a sum of percentages from different funds. It is neither a share of one pooled portfolio nor a market share. The reference is always our own coverage, not the fund market: a stock held by three funds here may well appear in many funds outside our catalogue. The history covers only the funds whose holdings reach back month by month, so each point also shows how many funds published a list on that date. A weight of 0.00 per cent in an issuer file does not mean the stock is absent; it means the position falls below the rounding threshold. We count such positions as held and leave them out of the weight.
Effective number of holdings
The number of positions says little about how a fund spreads its money: a fund with 1,200 holdings whose ten largest make up a third behaves differently from 1,200 equal positions. The effective number answers how many equally weighted holdings would produce the same spread. It is the reciprocal of the sum of the squared weights, which are normalised to 100 per cent first, because issuer holdings files do not always add up. Cash lines, futures and placeholders stay out, as do positions with a weight of 0.00 per cent, which do not change the spread. The figure describes the distribution of weights on the stated date and nothing else: it says nothing about volatility, default risk or the quality of the companies held.
Market network
The network draws only relations derivable from the holdings we track: which fund holds which stock, which index it tracks, which names an index contains, the sector and country a stock is classified under, how strongly two funds overlap by weight, and which stocks are held through largely the same funds. What is drawn is a section, not the whole: every ETF, index, sector and country we cover, plus the stocks with the widest fund coverage — the remaining names carry their own section on their detail page. Circle size states the number of connections within the drawn section and says nothing about size, quality or return. Because our fund coverage is unevenly distributed across providers, so is the network: a stock without connections is a stock without recorded fund allocations, not one nobody holds.
Overlap between two ETFs
For every holding that appears in both funds, the smaller of the two weights enters the sum, and that sum is divided by the smaller of the two holdings lists we track. The result is the share of capital financing the same company in both funds. The number of shared names is deliberately left out: it sounds larger and says less, because the many small positions of a world fund barely add up. The base is the holdings we have recorded for both funds, not a full hundred percent, and positions without a resolved identity are missing on both sides alike. The figure measures duplication in the holdings, not risk: two funds without overlap can still move almost identically.
Prices, performance and risk
Price charts
The chart shows closing prices, not the intraday path — highs and lows within a day are not visible in it. Dividends and distributions are not included, so the line is a price line, not a total-return line. Comparison series in the chart are indexed to the starting point so that instruments at different price levels can be compared.
US ETFs: holdings and costs
US funds report their positions for each quarter-end month on Form N-PORT, and the filing becomes public about 60 days after the reporting date. Weights refer to net assets. Cash, derivatives and collateral from securities lending are part of the filing, so the positions do not always add up to exactly 100 percent. The list of the largest positions shows stocks only. The expense ratio comes from the prospectus, or from the annual report for the gold trust. These funds are not counted in the ETF figures on our stock pages or in index compositions.
Performance tables
The periods run backwards from the most recent trading day; the starting value is the first quote on or after the cut-off date, not an interpolated one. Calendar years are formed from the last price of each year — the current year and the first year on record are therefore incomplete. Past performance says nothing about future performance.
Risk metrics
The maximum drawdown is the largest fall from a peak to the following trough, measured on closing prices. Volatility is the standard deviation of daily returns, annualized over 252 trading days. Best and worst year refer to calendar years derived from year-end prices; the shorter the history, the less these three figures carry.
Correlations
The value r is the correlation of two instruments' daily returns over 90 trading days: +1 means they move together completely, 0 means no relationship, −1 means they move exactly opposite. The comparison spans stocks, indices, crypto, commodities and FX, always on the same trading days. Correlation is not causation — and it is not stable: in stressed markets, otherwise independent assets often move together.
Returns in euros
For assets quoted in US dollars, we convert each daily close into euros at the European Central Bank reference rate for the same day. If that rate is missing (an ECB holiday), the most recent earlier rate applies, going back no more than five days. Year to date and one year follow the same rules as the dollar figures on the page. The difference between the two is the exchange rate move; bank spreads on conversion and taxes are not included.
Distance from peak, beta and correlation
The distance from peak compares the latest close with the highest close since our series begins, not with an earlier all-time high. Beta is the slope of a regression of the stock's daily returns on those of an exchange-traded fund tracking the S&P 500, over the last 252 shared trading days; the correlation measures how close that relationship was. Both change with market conditions and say nothing about how the stock will behave in the future.
Companies and funds
Company profile
The profiles describe lasting features of a company: its businesses, structure, headquarters and listing. They contain no figures that go stale quickly and no assessment. They are based on the annual reports and required filings the company submits to the US Securities and Exchange Commission and are written in our own words. The as-of date names the month in which the profiles were last reviewed.
Data as of:
Key financials
Revenue, earnings and balance sheet items are taken unchanged from the annual filing; margins, return on equity, equity ratio and growth are calculated from them. Only companies that file with the SEC are covered — which is why the block is missing for stocks outside the US. Fiscal years do not all end in December, so comparing two companies may compare different periods.
Stocks with similar ETF exposure
The list compares which of the ETFs we track hold each stock and scores the match as the share of jointly held funds among all funds involved (Jaccard similarity). Two names appear together because a portfolio reaches them through largely the same funds — not because their businesses are comparable. The reference set is our own fund coverage, which is skewed by provider; stocks with few fund allocations therefore get neighbours with a weaker match. The list is not meant as a selection for a portfolio.
Weight in an index fund since 2019
The chart shows the stock's weight in a fund for which we have holdings lists going back to September 2019. If several such funds hold the stock, we use the one with the longest record and, on a tie, the broadest index. Matching uses the ticker and the company name, so after a ticker change the line may start later. The weight reflects the fund, not the index itself: cash holdings and sampling shift it slightly.
Fund concentration
The share of the ten largest positions is measured against the sum of all positive weights; cash and derivatives with negative weights are left out. The effective number of holdings is 1 divided by the sum of the squared shares (the inverse of the Herfindahl index): a fund with 1,200 positions and an effective count of 90 is as diversified as one holding 90 equal positions. The table lists the last date in each year.
Changes in holdings
Added and removed mean the name appears in only one of the two lists, so a rename or ticker change can show up as both a removal and an addition. Weights also shift without any trading by the fund, because the holdings' prices move differently; a higher weight is therefore no sign of buying.
Cost example
The comparison is with the fund that has the lowest total expense ratio (TER) on the same index among the funds we track. The formula is end value = amount × ((1 + 5%) × (1 − TER)) to the power of the years, rounded to €10. Tracking difference, transaction costs, spreads, order fees and taxes are not included, and funds on the same index also differ in replication and distribution policy.
ETF coverage
The count is how many of the ETFs we cover hold a stock — not how many ETFs worldwide do. Our fund coverage is unevenly distributed across providers, so extrapolating to the whole market would be wrong. The percentile is more robust because every covered stock is measured against the same set: it gives the share of stocks held in strictly fewer ETFs. Where stocks tie, all of them receive the lower value. A high value says something about how widely a stock is held in index funds, nothing about the quality or valuation of the company.
Fund holdings
Funds publish their holdings with a delay and at differing intervals, which is why the as-of date sits directly at the block and is usually older than the prices on the same page. Weights refer to that date and shift afterwards through price moves alone. Sector and country breakdowns are computed from the same holdings.
Index composition
Index providers do not publish their full constituent lists freely. We therefore show the holdings of a large ETF on the same index — including that fund's own deviations: sampling instead of full replication, cash positions and an as-of date in the past. That is enough for orders of magnitude and concentrations, but it is not an official index list.
Published fund performance
We do not calculate these figures ourselves. They are the published figures for the fund and follow its conventions — usually including reinvested distributions and net of fund costs. That is why they differ from our own performance table on the same page, which is derived from prices. Cut-off dates and currency may differ as well.
Crypto, macro and tools
Market capitalization and rank
The rank follows from market capitalization, that is circulating supply multiplied by price. How many units actually circulate is in part self-reported by the projects, and locked or lost holdings are not treated consistently. The rank therefore moves even without a large price change.
Altcoin Season Index
The index counts how many of the largest altcoins outperformed Bitcoin over the last 90 days and expresses that as a percentage. It is based on the leading coins by market capitalization with a sufficiently long price history — stablecoins and coins without history drop out. The index measures relative strength against Bitcoin, not the direction of the crypto market as a whole.
Country indicators
The series are annual and published with a substantial lag; the value shown is the most recent available one, not the current year's. Not every country reports every series, so gaps stay visible rather than being estimated. Stock and ETF lists on the same page come from our own coverage and are not a complete market overview.
Real interest rate
Euro area: ECB deposit facility rate minus euro area HICP inflation. US: effective federal funds rate minus CPI inflation. Both series are taken unchanged from the official publications. The inflation rate looks back a year while the policy rate applies today; a more exact real rate would subtract expected inflation, which cannot be observed directly.
Economic calendar
Two kinds of entries: releases with a computable rhythm, such as "third business day", come from the rule. The US jobs report instead carries the dates of the official release schedule published by the US Bureau of Labor Statistics, because no rule of thumb gets it right — holidays, reference weeks and funding lapses move it. For a month with no scheduled date on file the calendar shows nothing rather than estimating one. Central bank rate decisions remain absent: those meeting dates are set by the committees. All times are computed in UTC and converted only for display; a change the publishing body announces at short notice is not reflected.
Model portfolios
The weights are fixed in advance and applied retrospectively to the price series of the building blocks. Such a backtest already knows the outcome and leaves out transaction costs, taxes, spreads and one's own behavior during drawdowns. It shows how an allocation would have behaved — not how it will behave.
Calculators and tools
The tools compute deterministically from what is entered — different assumptions produce different results, and none of them is more likely than another. Transaction costs, spreads, taxes and exchange rates are included only where the individual tool asks for them explicitly. Historical price series only cover what actually happened; rare events are under-represented in them.
Assets: auction prices, property indices and comparison
Every dot is a price actually paid at auction, taken from public auction archives — not an estimate and not an asking price; prices include the buyer's premium, so they reflect what a buyer really pays. Because collectibles do not trade daily, the line is the median per period — monthly for positions with enough sales, quarterly otherwise; individual sales stay visible as dots so the spread is not hidden behind the median. In the comparison against the S&P 500, MSCI World, Bitcoin and gold, all series are indexed to 100 at the start of the common period and drawn on a logarithmic scale, because a four-digit return and a two-digit one do not fit the same linear axis. The period follows the shortest series involved; correlations are shown only where at least three years of common history exist. Residential property is the exception: there are no individual sales, only official price indices from Eurostat and the US Case-Shiller series. They track a typical property, are nominal and exclude transaction costs, rental income and financing. For Pokémon, print runs are separate positions: 1st Edition, Shadowless and Unlimited of the same card trade at different prices, and merged into one series the line would track the period's mix of print runs rather than the market. For watches and cars, special variants are excluded — a Paul Newman dial or a Turbo S Leichtbau is its own market and would single-handedly set the median of a thinly traded period.
Price index for a collectible asset class
A period median consists of different items sold in each quarter. If it rises, that may be the market — or simply that more expensive pieces happened to be offered; with collectibles, whose prices differ by orders of magnitude, this composition effect is not a marginal issue. The index avoids it: only the price change of the same standardized good between two sales is evaluated, estimated by least squares following the method of Bailey, Muth and Nourse (1963) — the same one behind official repeat-sales indices for residential property. The size of the difference shows in the cross-check made when building it: for trading cards the index reaches a factor of 5.45 over the same period, an independently computed equal-weighted mean of the change per card reaches 5.49 — while the raw class median reaches 9.55. The index is computed only where the assumption holds. Two PSA 10 copies of the same card are interchangeable through the grade; two cars of the same generation are not, and watches of the same reference differ considerably in condition, original parts and papers. For vehicles, watches and residential property no index is therefore computed, with the reason stated on each page. What the method does not do: positions never sold a second time do not enter, frequently traded ones are overweighted, and pairs less than 30 days apart or more than a factor of 20 apart are discarded — those are reversals and mis-assignments, not market moves.
Ranked by return per year
Only positions with at least ten documented sales are included. Without that threshold, a position with five sales would top the list, half of whose return comes from a single hammer price. The return is the annual change between the first and the most recent period value; periods shorter than a year are not annualized. The real return adjusts for average inflation over the same period, computed the Fisher way rather than as a difference — with double-digit returns simple subtraction is noticeably off. What governs is the currency of the price: hammer prices are in US dollars and are adjusted with US inflation, property indices with that of their own currency area. If the inflation series covers less than four fifths of the period, the real figure is omitted rather than treating the missing months as inflation-free. Next to the ranking stands the median across all covered positions: a list of winners alone describes the selection, not the asset class. Sales per year and the typical gap between two sales tell you how long you may wait for a buyer — the worst-documented property of this asset class. All returns are gross price changes; premium, commission, grading, storage and shipping costs are not deducted.
How real assets and markets move together
The calculation runs on logarithmic quarterly changes. Quarters are the finest cadence real assets support: auctions do not happen daily, and a monthly value per class would sometimes rest on a handful of hammer prices. Each pair gets its own window because the series begin at very different times — a single common window would cut everything down to the shortest series. That is why the number of shared quarters stands on every value; below twelve, nothing is shown. Class series are built differently, and the table says so: a matched-model index for trading cards, the quarterly median of all hammer prices for vehicles, the equal-weighted mean of official indices for residential property. A median measures market and selection at once, a matched-model index only the market — the rows are therefore not equally dependable. On the equity side, index futures serve as the benchmark instead of ETF prices: the latter only reach back to mid-2024 and would yield eight quarters, fewer than the minimum overlap. The S&P 500 and Nasdaq 100 sit in the matrix as a sanity check — two equity indices must be well above +0.8, otherwise the calculation is wrong. Relationships are not causes, and a value near zero does not mean independence, only no linear co-movement over the period measured.
Book data and purchase links
The English catalog is curated by hand, one entry per book, and lists standard works from many publishers rather than one publisher's programme. The factual details — title, subtitle, author, ISBN, format, page count, publication year — are checked against at least one source independent of the retailer, and a page count or year we cannot confirm is left out rather than guessed. No price is shown: outside Germany's fixed book price there is no publisher's price to state, and retailer prices we are not permitted to display. No sentence of the descriptions is taken from a publisher's or retailer's blurb; they state our own assessment and say where a book is dated, narrow or written by someone selling something. Not every book is assessed — titles not yet published, or ones we have no dependable basis for, are listed with their facts and no description. Cover images are shown as product illustrations for the linked book and served from our own domain rather than hotlinked; the rights remain with the publishers. Purchase links go through the Amazon affiliate programme and open Amazon's German store in English, because that is the marketplace our affiliate account belongs to: we receive a commission on a purchase, the price for the buyer is unchanged, and neither the selection of titles nor their assessment depends on it. Each title additionally carries a score from 3 to 5 and a place in a ranking across the whole catalog. The scale starts at 3 because books we consider weak are not listed at all — 3 means useful, not poor. Whether a title scores 3 or 4 is decided by the build from traits recorded in the catalog and open to checking: number of printings, extent, publication year, the depth of our own assessment, with deductions for unpublished and unassessed titles. The same catalog therefore always yields the same score. A 5 is not a calculation but a hand decision: it is written into the catalog and reserved for titles that carry their subject over years. The ranking follows from score, book-pick flag and the same traits; it orders, it does not measure. Titles marked as an Alysly Book Pick are ones Alexander Bock has read and recommends; their rating is marked up as a review by a named person, while the remaining assessments are marked up as editorial work. That selection does not depend on the affiliate programme either. Titles without an own assessment do carry a score derived from the facts, but are not marked up as a review, because the read basis is missing.