How the figures are built

For each market, we add up the dollar volume of the companies we track there — the number of shares that changed hands in the session, times the price — and convert it to US dollars at that day's exchange rate. The result is one figure per market per session: how much money changed hands in the companies we follow.

The United States is the exception: its figures are the whole market. Every stock and every exchange-traded fund listed on a US exchange, from our data provider's end-of-day figures, shown as three lines — US stocks, US ETFs, and the two together — in a group of their own, ahead of the other markets. Checked against the total reported across every US exchange and trade-reporting facility, it came to between 99.7% and 100.3% on each of six sessions in September 2026.

A day counts as a session only when every company we track in that market traded. When one exchange in a market is closed and another is open — a holiday on the mainland while Hong Kong trades, for example — the market shows its most recent full session, with its date, rather than a total for part of the market.

Companies tracked in each market at the latest session: 169 in China, 110 in Japan, 76 in the UK, 74 in Canada, 67 in Brazil, 65 in Germany, 64 in Israel, 58 in South Korea, 57 in France, 48 in Italy, 42 in Spain, 40 in the Netherlands.

Exchanges and index providers report volume in different units. Some report the number of shares traded; some report the value traded in the local currency; some report turnover for an index. A page can show an index level beside a figure labelled “volume” and leave you unable to tell whether that volume is a count of shares or an amount of money — and neither figure says how much money actually changed hands.

Figures built in different units can't be compared with each other: a share count in one market and a value in euros in another are not two measurements of the same thing. And a figure in a local currency can't be compared with itself over time once the currency moves, because part of any change is the exchange rate rather than the trading.

A session total is built the same way in every market and on every day: dollar volume of the same set of companies, converted at that day's rate. That is what makes a figure for Japan comparable with one for Germany, and this week's comparable with last month's.

The US figures are the exception here too. Being the whole market, they are not comparable with another market's tracked companies — which is why they sit in a group of their own rather than in the same row. Each is still comparable with its own history.

Outside the US, it is not the volume of the whole market: each figure is the companies we track there, not everything that traded, which is considerably more. Those figures compare because they are built the same way, not because they cover everything.

In the US, the market's figures and our company pages are two different things. The totals are the whole market; the pages cover the companies we track. And the reading of the US in the events — a session “across tracked US companies” — is those companies' own ratio to their normal, not the market's: on the same day the two can differ, and neither is wrong.

Each total is shown against the market's own normal: the average of its previous 20 sessions. “1.50× normal” means the session traded half as much again as that average. The comparison is always with the same market's own history, never with another market's.

Whole-US figures for sessions up to late September 2026 were rebuilt afterwards from a single fetch of that day's listings, and each is marked as rebuilt on the Tape. They leave out companies delisted since, and understate the oldest sessions by up to about one and a half per cent. Their average — the normal the first US readings were measured against — is very slightly low, so those readings are marginally high, by less than the rounding for most sessions; the effect fades as measured sessions replace rebuilt ones.

It needs 20 earlier sessions, and it is taken over the companies that traded on each of them, so a company that joined the tracked set recently does not make the market look busier than it was. Until a market has that history, its total is shown on its own, without a comparison.

Turnover is a company's dollar volume in a session divided by its market capitalisation: the share of the company that changed hands that day. It is the measure the platform is built around.

It allows for size where a dollar figure can't. A billion dollars of trading is a great deal for a company worth twenty billion and routine for one worth three trillion; as a share of each company's size, the two sessions can be set side by side. It does not allow for how much of a company is held still (below).

Market capitalisation is worked out for each session: that day's closing price times the company's shares outstanding. Share counts are refreshed monthly — for US companies from the count on the cover of the latest quarterly or annual filing with the SEC, when that filing is recent; otherwise from the data provider. A new count that would move a company's market capitalisation by more than half is not used until a person has checked it: the previous count is kept meanwhile. Turnover is that session's dollar volume divided by this market capitalisation, one figure everywhere on the site. Prices are adjusted for stock splits, not for dividends, so a past session's price is the one it actually traded at, allowing for any split since. Because the same adjusted price sits in both the dollar volume and the market capitalisation, the adjustment cancels out of turnover.

Market value of companies with more than one class of shares. A company is valued once, as the sum of its share classes. This applies to companies whose share counts come from SEC filings; companies in other markets are described below as their rule is set. A class is counted if it has the same dividends per share as a listed class and an equal share of what remains on liquidation, or if it converts or exchanges into one. A listed class is valued at its own closing price. Any other counted class is valued at the closing price of the class it converts or exchanges into, at the rate the company's filing states. Classes that carry votes only are left out, and so are shares a company holds in its own treasury. Some companies are organised as an “Up-C”: the public company's shares sit alongside units in an operating partnership, which their holders can exchange for listed shares. There, the units are counted as if exchanged, and the voting-only shares paired with them are not. A listed line's turnover is its dollar volume divided by the company's value that session, the market cap shown beside it (from 15 October 2026; before that, by the value of its own class), so the table, the company page and the insights show one figure. Share counts come from the cover of each quarterly or annual SEC filing. A company that files only an annual report (Form 20-F) is counted from that report, and its date is shown. From 10 October 2026, every history is on this basis: each session uses the share counts, conversion rates and treasury figures of the filing in force on that day. Events already published in the feed are left as they were published; findings on company pages are recomputed on this basis.

Turnover is measured against all shares outstanding, not the free float: the shares actually available to trade. Where much of a company's stock never trades, this understates how actively it is traded, by a factor of one over the share that does trade. A company with a fifth of its shares available reads five times quieter than one whose shares are all available. The cases are known and large: state holdings in Chinese companies and cross-shareholdings between Japanese companies.

So read a company's turnover against its own history, which is what the platform's comparisons do: a holding that does not change cancels out of “times its normal” and of where a session sits in its own year. Comparing turnover levels between companies, and above all between markets, mixes in how much of each company is held still. One ranking does compare levels across markets: the day's findings put first any session that turned over more than 20% of a company's value (or moved its price more than 20%). A company with little of its stock in circulation needs proportionally more trading to clear that line.

Turnover compares across companies because it is a percentage of each company's value. An event that calls a session heavy makes a narrower claim: that it was among the heaviest 1% of sessions for the companies we track in that market, over the last six months — and, where a market has enough companies to group, among companies of its size. So 0.5% can be heavy in France and ordinary in the US: each market is measured against itself, and the event states the mark it was measured against.

Companies worth under $25 billion routinely trade several percent of their value in a day. For them, and in a market where we track too few larger companies to measure a mark, the comparison is with the company's own past year instead: among its own heaviest 1% of sessions. The two use different windows on purpose — six months for a market, because markets shift from one period of trading to another, and a year for a company, because a single company needs longer to show what is normal for it.

A company that trades heavily as a habit is called heavy at most once in ten sessions, unless a session is half as heavy again as the last one. And a company is not called anything until we have 120 sessions of its history: before that, it would be judged mostly by its market rather than by itself (see the companies currently affected under Known limitations).

In some markets a few companies trade far more heavily than the rest — in the US, among the largest companies, two set the mark for all of them. A company whose usual session is under a tenth of its market's mark is compared with its own past year instead, and its event says so. This is checked every day, so a company can move between the two comparisons as its trading changes.

Two labels on the site are our own, and they describe different things.

The Infinity cycle counts each market's trading sessions in threes, from that market's own calendar, so every session is Day 1, 2 or 3 (“Day 1 of 3 · Infinity cycle”). It is a way of reading the calendar, not a signal: measured over 504 sessions in every market, the three days do not differ in trading volume or in the size of price moves.

The Infinity sectors group US companies into 29 sectors by what they sell, drawn more finely where capital is concentrated (six separate AI buckets) and more broadly where it is not. Where a company's business spans sectors, its revenue is split across up to four of them; most companies have one. The classification is our editorial judgement, and companies added in September 2026 were classified from their own descriptions of their business in their SEC filings. A company's label is its largest sector. Sector totals split each company's dollar volume by its mix, so they add up to the market's total.

Every sector view says what it rests on: how many companies it counts, and their share of the trading in that market's index (the FTSE 100, the Nikkei 225; the S&P 500 in the US). The share is of the index's trading, not the whole market's. In the US, the same companies were 44.8% of the trading in every stock listed on a US exchange on 2026-10-23.

Outside the US, a company is classified from its own annual report where one can be read: European companies' reports through the filings index at filings.xbrl.org, and Japanese companies' annual securities reports (有価証券報告書) from EDINET, the Financial Services Agency's disclosure system. Where no filing serves, the classification rests on a dated revision of the company's Wikipedia article, read but not reproduced. A quotation not in English is shown as written, with our translation beneath it.

出典：EDINET閲覧（提出）サイト（https://disclosure2.edinet-fsa.go.jp/）、PDL1.0. Quotations from Japanese filings are excerpted and translated by InfinityVolume; the translations are ours, not the companies'.

Changed on 29 September 2026. Aerospace became a sector of its own, taking GE Aerospace's commercial engine business, Boeing, RTX, Howmet and Honeywell Aerospace from Industrial Capital Goods & Transport and from Space, Defense & National Security. Lululemon, McDonald's, Marriott, Hilton, Royal Caribbean and Carnival moved from Commerce & Consumer Platforms to Consumer Staples & Brands, under a rule that classifies a consumer company by whether it owns the brand it sells rather than by how it sells it. Corning moved from AI Systems, Networking & Optics to Materials & Critical Inputs, where most of its sales are.

Sector totals before and after that date are not comparable — Consumer Staples & Brands now carries roughly half as much volume again as it did, and in some sessions nearly double. Market totals are unaffected: the changes move volume between sectors without adding or removing any.

Changed on 15 October 2026. From 15 October 2026, a sector's share is of all covered companies' dollar volume in its market, the US extended tier included, where it was of the tracked companies' alone, and a sector's total counts every classified company we cover, the extended tier's included. Companies we have not yet classified are shown as Unclassified, so the sectors and Unclassified together make up all of it; the Unclassified share shrinks as classification proceeds. In the US the sector figures change once on that date; sessions before it are not comparable with sessions after it. Outside the US the covered companies are the tracked companies, so nothing changes there.

A member's company page splits the company's weekly share-price moves over the past 52 weeks into parts that add up to 100%. Each part is the share of the weekly ups and downs that one more influence explains, after the ones before it: first the overall market, the company's home market (below); then its most correlated stocks, the tracked companies whose weekly moves follow its own most closely, in any sector; then, where one stands out clearly, one wider factor; the rest is other things, such as company news. A part under half a percent reads “under 1%” and counts as zero. Weekly returns are in the company's own currency.

The home market is measured through a fund listed on that market that tracks its main index (for a company listed in Hong Kong, the Hang Seng), in the market's own currency, over the same weeks as the company: the US, the S&P 500 (SPY); Japan, the TOPIX (1306.T); China (Shanghai and Shenzhen), the CSI 300 (510300.SS); China (Hong Kong listings), the Hang Seng (2800.HK); Korea, the KOSPI 200 (069500.KS); the UK, the FTSE 100 (ISF.L); Germany, the DAX (EXS1.DE); France, the CAC 40 (CAC.PA); Italy, the FTSE MIB (FTSEMIB.MI); Spain, the IBEX 35 (LYXIB.MC); the Netherlands, the AEX (IAEX.AS); Canada, the S&P/TSX 60 (XIU.TO); Brazil, the IBrX 100 (BRAX11.SA); Israel, the TA-125 (TCH-F2.TA). The card names the index. A company that makes up more than 5% of its home market but whose market part is under 5% is said to have moved largely independently of its home market that year; its weight is approximated by its share of the combined market value of the companies we cover there (Hong Kong listings and Shanghai/Shenzhen shares apart), since we hold no index provider's constituent weights.

Peers are up to three companies we track, in any sector or market, whose weekly moves followed the company's most closely over the year before, and still did over the past year (a correlation of at least 0.5 in both). They count equally. A company's own other listings and share classes, and its holding companies, subsidiaries and large affiliates, are never its peers. A company with less than two years of history has its peers chosen on the same year, and the card says so.

The wider factor is one of six: the 10-year US Treasury yield's weekly change, oil (the USO fund), gold (GLD), the US dollar (UUP), bitcoin (IBIT), and growth stocks over value stocks (the IWF fund less the IWD fund). The one that adds the most is shown only when it adds at least 5 points and passes a statistical test clearly enough to rule out having been picked by chance from six; otherwise the card says no clear extra factor.

Earnings weeks (US companies only, from their results filings with the SEC) are shown beside the bar: the share of the year's movement that came in those weeks. Heavy days are sessions whose dollar volume was at least 1.5 times its average over the previous 20 sessions; the card gives the same-day price move on the heaviest ten of the past year. None of this is a forecast.

The 10-year yield: Board of Governors of the Federal Reserve System (US), Market Yield on U.S. Treasury Securities at 10-Year Constant Maturity, Quoted on an Investment Basis [DGS10], retrieved from FRED, Federal Reserve Bank of St. Louis; fred.stlouisfed.org/series/DGS10.

Every figure on the platform describes sessions that have happened: how much traded, how that compares with normal, how much of a company changed hands. None of them is a forecast, and none should be read as a signal of what the next session will do.

Prices and volumes for US-listed companies, and the whole-US-market totals, come from Tiingo's end-of-day data. Shanghai and Shenzhen listings come from Tiingo or our international end-of-day data provider, company by company. Hong Kong listings, ETFs and every other market come from that provider. Currencies are converted at the European Central Bank's daily reference rates. Share counts come from filings with the SEC for US companies when the filing is recent, and otherwise from the data provider. Treasury yields, credit spreads and economic indicators come from the Federal Reserve Bank of St. Louis (FRED). GDP figures, used for market-capitalisation-to-GDP ratios, come from the World Bank.

A member's live price on a US company page is Tiingo's derived reference price, consolidated from validated trades and quotes across several trading venues, not one exchange's last sale. It is shown beside the close and is never used in any figure.

The figures are updated four times each weekday, after the Asian, European and US closes and again late in the US evening. A market's figure carries a date whenever its latest full session is not today's.

Market Data from Tiingo.com

Logos provided by Logo.dev

Share counts lag. They are refreshed monthly, and an SEC count reflects the date of the filing it comes from. Shares issued or bought back since then show up at the next refresh; until they do, that company's market capitalisation and turnover are off by the change.

One exchange rate a day. Each day's figures are converted at that day's reference rate, not the rate at the moment each market closed.

The figures follow the provider's data, including its errors. Where one company's volume is missing for a day its market traded, we record it as unknown rather than zero, and it is left out of averages and totals.

Found on 25 September 2026 and corrected from 26 September: closed days stored as sessions. Until then, days on which some exchanges, Tel Aviv among them, were closed were stored as trading sessions with no volume. They lowered those markets' 20-session averages, so their “× normal” figures read high, and they shifted some trading-cycle labels. Cycle labels already published for the affected markets on those dates were left as published.

New listings are silent until they have 120 sessions. A company with a short history has no events at all, by decision: judged mostly by its market, it would be called unusual session after session. As of 26 September 2026 this applies to Cerebras (until about 3 November 2026), SpaceX (about 2 December 2026), CXMT (about 19 January 2027) and Unitree (about 18 February 2027). No events for one of these companies means this, not that nothing happened.