The favorite scrolling headlines of market apps and financial news: “northbound capital net inflow of ¥8 billion today,” “main capital net outflow of ¥500 million,” “such-and-such stock hit the dragon-tiger list,” “margin balance at a record high.” This is the finale of the A-Share Concepts Explained series — we’ll explain what each of these things is, and more importantly, how much water each one’s statistical methodology holds.
Northbound flows: money coming in through Hong Kong
Stock Connect （沪深港通） is the channel linking the two markets: overseas investors buy and sell A-shares through the Hong Kong exchange, and that direction is called northbound (flowing north into the mainland); mainland investors using it to buy Hong Kong stocks is southbound. The main bodies behind northbound flows are overseas institutions — sovereign funds, pension funds, hedge funds, and passive index money (the allocation flows that came with MSCI’s inclusion of A-shares) — which is why the market watches them as “smart money.”
Two recent changes you must know:
- Real-time data has been discontinued. Since August 2024, the Hong Kong exchange no longer discloses real-time intraday net buying by northbound capital — only low-frequency data such as total holdings remains. The old “northbound real-time inflow curve” in market apps is gone; anything now claiming to show “real-time northbound” is an estimate or stale news.
- Northbound was never a single entity: it’s the aggregate of thousands of overseas institutions, including long-term allocation money and fast-in-fast-out hedge funds. Imagining it as a single well-informed “foreign-capital brain” with a unified opinion was a misreading from the start.
“Main capital net inflow”: ground zero for methodology traps
This is the most widely circulated — and most misleading — data in market apps, bar none. First, how it’s computed: the app classifies trades by order size — say, single orders above ¥1 million count as “main/extra-large orders,” ¥200k–1M as “large orders,” and anything below as “retail orders” — then subtracts “main” sell-side turnover from buy-side turnover to get “main net inflow.”
The methodology makes distortion almost inevitable:
- Big money can be chopped small: one ¥100 million sell order split into 500 orders of ¥200k registers in the statistics as “retail selling,” and “main net outflow” instantly becomes “retail net outflow.” Order-splitting is a standard feature of every trading platform, and doing it costs nothing.
- Retail orders can register as big: in a bull market, a few wealthy individuals buying the same stock can each place orders well over a million, and they get recorded as “main capital.”
- Inflow and outflow are always equal: every trade has a buyer and a seller; “net inflow” is only a classification difference by order size, not actual money entering or leaving the stock.
The conclusion, stated directly: “main capital net inflow/outflow” provides no basis whatsoever for any decision. Its only honest use is describing “were big orders or small orders more active today” — and volume itself already tells you that.
The dragon-tiger list: the exchange’s official “abnormal movement” bulletin
The dragon-tiger list （龙虎榜） is a daily post-close exchange publication naming the top five brokerage branches on the buy and sell sides of abnormally moving stocks. The listing criteria are hard-coded in exchange rules; common ones include: daily price-move deviation reaching ±7%, daily turnover rate reaching 20%, cumulative deviation of ±20% over three consecutive trading days, and so on.
The list’s value is that it’s real trade data (not an estimate): you can see the five brokerage branch seats with the largest buy amounts. The famous “hot-money seats” of trading lore come from here — certain branches became notorious for appearing constantly with a ferocious style (“such-and-such securities, such-and-such road branch”), and retail circles have even formed a school of “following hot-money seats.”
The cold water, as always: a seat ≠ a person. Once a famous hot-money seat is exposed, switching seats and splitting positions across multiple seats is routine; and appearing on the list means the move has already happened — you’re looking at a result, not a forecast. Using the dragon-tiger list as a heat map of “which stocks are being gamed right now” is fine; using it as a copy-trading signal is not.
Margin trading: borrowing money to buy, borrowing shares to short
Margin trading and securities lending （两融） are the credit services brokers provide:
- Margin buying （融资）: borrow money from the broker to buy stocks, sell after a rise and repay; both gains and losses are amplified by leverage. The margin balance (the market’s total outstanding margin debt) is often used as a thermometer of sentiment — a rapidly climbing balance means more people are borrowing to bet on rises.
- Securities lending （融券）: borrow shares from the broker and sell them, buy them back after a fall and return them, profiting from the decline. A-share lending supply has always been scarce, and in 2024 regulators tightened sharply (relending of borrowed shares suspended, margin ratios raised), so the actual short-selling scale is small. A-shares are basically a one-sided market, and “short-selling pressure” is more a psychological concept than a real force.
Margin trading has a ¥500,000 asset threshold, with interest around 6%–8% annualized. Our advice for beginners matches the first series: leverage doesn’t create returns, it only amplifies volatility — this is a door you don’t need to open within your first two years.
Finale: four principles for beginners reading data
At this point the series has touched essentially every concept commonly seen in market apps. We’ll leave you with a principles checklist for reading data — it applies to any new indicator beyond this series too:
- Ask about the methodology first: how is this figure compiled? What’s the denominator? Where are the classification boundaries? Data with an unclear methodology is as good as no data (“main capital” is the perfect counterexample).
- Distinguish commitment from statement: executed trades are commitments backed by real money; resting orders, opinion polls, and “bullish-percentage” readings are zero-cost statements. Always weight the former.
- Lagging is the norm: every piece of public data you see — the dragon-tiger list, quarterly reports, northbound holdings — comes after the action itself. The value of public data is verifying a thesis, not front-running.
- No single indicator makes a decision: Part 1’s turnover rate must be read alongside the price position, Part 5’s P/E alongside the industry, and this article’s dragon-tiger list alongside seat history. Any product claiming “one number tells you when to buy and sell” is selling anxiety.
The next two series are already on the way: A-Share Quant in Practice: From Data to Live Trading (placeholder reserved) will turn these concepts into backtestable strategies — rules like “hold the ETF above its moving average,” validated against ten years of data to see whether they actually work. The concepts series is now complete.
Disclosure rules and regulatory policies change; refer to the exchanges’ latest announcements. This article does not constitute investment advice.