What it measures. Won that retail investors have borrowed from brokers to buy stock. When it stops rising and starts falling, the leverage build-up is being worked off.

Formula. The margin balance itself, in ₩T. Reported in millions of won; converted ×1e-6.

Reference levels. Pre-hype normal is ₩16–20T (calm median 19.0, 90th percentile 23.7). Stable at or below ₩20T, neutral to ₩25.6T — the 2021 mania peak — red above it.

Caveat. A level says how much leverage is outstanding, not whether it is unwinding voluntarily. Read it beside forced liquidation: the same balance falling while that indicator is red means brokers are selling collateral.

What it measures. The share of overdue margin positions that brokers auto-sold that day. Spikes mark a margin-call cascade.

Formula. Today's ratio, with no averaging. Stable below 1.1, neutral below 3.0, red at or above 3.0.

Reference levels. Calibrated on the daily readings since 2024: calm median 0.7, 85th percentile about 1.0–1.1, 95th about 3.1.

Caveat. The definition changed in late 2023 — uncollected receivables bloated the denominator and stepped the ratio down 5–10×. This series therefore starts in 2024, and earlier values are not comparable.

What it measures. How much money is actually moving through 2× long funds. The margin balance only sees credit extended by brokers; this is the leverage retail buys through the exchange, and it moves first.

Formula. Daily ₩ traded in KODEX Leverage (KOSPI 200 2×) plus KODEX KOSDAQ150 Leverage — closing price × share volume, summed.

Reference levels. Stable below ₩1.0T a day, neutral to ₩2.0T, red above. Calm years sit remarkably flat: 2021–2025 medians all fall between ₩0.44T and ₩0.69T.

Caveat. Turnover is churn, not holdings — the same won can trade many times a day, so it measures activity rather than the size of the pool. Inverse and 2× short funds are excluded: they are leveraged bets too, but bearish ones, and including them would make the number rise on both fear and greed.

What it measures. Margin debt measured against the cash sitting in broker accounts — how stretched leverage is, rather than how large it is.

Formula. Margin balance divided by customer deposits held at brokers, as a percentage.

Reference levels. Stable below 35%, neutral to 38.1%, red above. Korean desks call 35% burdensome and 40% the top; 35% is also the 63rd percentile of the stored record.

Caveat. Both sides scale with the market, so unlike the margin balance this does not drift as the index grows. It is an aggregate: if deposits swell with idle new accounts while debt concentrates among active traders, the ratio understates the risk — which is why the absolute balance is kept beside it.

What it measures. Net buying or selling by foreign investors — the flow that has driven the drawdown. A turn needs their selling to slow sharply or flip.

Formula. Net buying minus selling by foreign investors that day, in ₩T, as reported.

Reference levels. Stable when foreigners bought that day. Red below −₩1.8T, about the 1st percentile of calm-era days.

Caveat. A raw won figure, so it grows with the market: ₩1T of selling is less pressure at a KOSPI of 6,600 than at 2,500. It is shown unadjusted anyway, because a deflated number matches nothing any Korean source publishes and cannot be checked.

What it measures. The capital-flight tell. A genuine foreign exodus weakens the won quickly.

Formula. The USD/KRW rate itself.

Reference levels. Stable at or below 1,313 (the calm-era median), neutral to 1,429.5 (the 90th percentile), red above. A weaker won is the capital-flight tell, so higher is worse.

Caveat. The won also moves on rate differentials and dollar strength that have nothing to do with Korean equities.

What it measures. Whether a regional move is arriving in Korea amplified — the signature of leverage still in the system — or one-for-one.

Formula. Beta of KOSPI to the Nikkei over the last 5 sessions: cov(KOSPI, Nikkei) ÷ var(Nikkei) on daily returns.

Reference levels. Stable below 1.2, neutral to 1.8, red above. Beta near 1 means Korea moves with the region; 2 means a regional wobble arrives in Seoul at double size, which is what leverage does.

Caveat. Beta measures amplitude, not direction: Korea falling alone while Japan is flat produces a LOW beta and reads stable. A 5-session window is deliberately responsive and therefore noisy — read it as a current estimate, not a trend.

What it measures. The closest regional comparator. Korea and Taiwan are both semiconductor-heavy export economies, so Korea amplifying Taiwan means Korean leverage rather than a regional tech de-rating.

Formula. Beta of KOSPI to the TAIEX over the last 5 sessions: cov(KOSPI, TAIEX) ÷ var(TAIEX) on daily returns.

Reference levels. Stable below 1.2, neutral to 1.8, red above. Taiwan isolates the semiconductor cycle; Japan reads broader regional risk appetite.

Caveat. Same limits as the Japan beta: it reads amplitude rather than direction, and 5 sessions is a small sample. Taiwan also carries cross-strait and TSMC-specific risks unrelated to Korea.