Korean retail investors buy the dip.
Does betting against them pay?
Korea publishes, every trading day and for every stock, how many shares retail investors, foreigners and institutions each bought net. Every broker app shows a "top retail net buying" list, and every trader has heard the advice: do the opposite of retail. Using the same method as my foreign-flow article, I counted six years of it — this time for retail.
01The data and the rules
KOSPI is Korea's main board, KOSDAQ the secondary one (smaller, more volatile). The session runs 09:00–15:30 Korea Standard Time (UTC+9). Korea has no daily equivalent elsewhere of what it publishes after the close: per-stock net buying in shares by three investor categories — retail (individuals), foreigners and institutions.
Because the confirmed figure only exists after the close, nothing here buys on the signal day. Every trade is at the next morning's open.
- Retail net buy — any positive amount
- Retail heavy buy — net buying of at least 5% / 10% of that day's volume
- Retail top-20 net buy — the 20 largest retail net purchases by value that day within the universe
- Mirror images: retail net sell / heavy sell / top-20 net sell
Buy the next open, sell at the open 1 day, 1 week, 1 month or 3 months later; 0.22% deducted on every buy. The baseline is the same 749 stocks bought on any random day. The median is the middle result when every trade is lined up from worst to best — half did better, half worse.
02What kind of day does retail buy on?
Before following the signal, I checked what the stock did on the signal day itself.
| Signal | Count | Share of days closed up | Median |
|---|---|---|---|
| Any day (baseline) | 1.11M | 45.5% | 0.00% |
| Retail net buy | 566,731 | 27.6% | −0.86% |
| Retail heavy buy (10%+ of volume) | 4,533 | 9.4% | −1.52% |
| Retail top-20 net buy | 29,989 | 18.4% | −1.51% |
| Retail net sell | 539,351 | 65.1% | +0.78% |
| Retail heavy sell (10%+ of volume) | 3,802 | 89.7% | +2.00% |
| Retail top-20 net sell | 29,989 | 82.7% | +2.15% |
When retail buys heavily, it is a day the stock fell — nine times out of ten. When retail sells heavily, nine times out of ten the stock rose. Korean retail buys dips and sells rallies: averaging down and taking profits. Foreigners and institutions are the other side of those trades, which is why "foreigners buy on up days" looks like insight.
This explains everything that follows. "Top retail net buying" is another name for "the stocks that fell hardest today", and "top retail net selling" is "the stocks that rose hardest today".
03Following retail
| Signal | Hold | Win rate | Baseline | Median | Baseline |
|---|---|---|---|---|---|
| Retail net buy (any) | 1 week | 46.4% | 46.2% | −0.41% | −0.36% |
| 1 month | 47.9% | 47.5% | −0.52% | −0.46% | |
| 3 months | 48.8% | 48.4% | −0.51% | −0.45% | |
| Retail heavy buy (10%+) | 1 week | 42.8% | 46.2% | −0.61% | −0.36% |
| 1 month | 44.5% | 47.5% | −0.94% | −0.46% | |
| 3 months | 43.6% | 48.4% | −1.99% | −0.45% | |
| Retail top-20 net buy | 1 week | 46.8% | 46.2% | −0.41% | −0.36% |
| 1 month | 47.3% | 47.5% | −0.70% | −0.46% | |
| 3 months | 47.1% | 48.4% | −1.42% | −0.45% |
"Retail bought" on its own carries no information: the numbers match the baseline to the decimal. Retail is a net buyer on half of all stock-days, so that is no surprise.
Stocks retail bought heavily did badly: 43.6% winners after three months, median −2.0%, against 48.4% and −0.45% for a random purchase. So far the advice — avoid the top retail buys — holds.
04So do the opposite?
Here is the point of the article. If heavy retail buying is bad, heavy retail selling should be good.
| Signal | Hold | Win rate | Baseline | Median | Baseline |
|---|---|---|---|---|---|
| Retail heavy sell (10%+) | 1 week | 44.1% | 46.2% | −0.61% | −0.36% |
| 1 month | 45.5% | 47.5% | −0.91% | −0.46% | |
| 3 months | 43.6% | 48.4% | −2.13% | −0.45% | |
| Retail top-20 net sell | 1 week | 45.7% | 46.2% | −0.58% | −0.36% |
| 1 month | 46.4% | 47.5% | −1.01% | −0.46% | |
| 3 months | 46.5% | 48.4% | −1.67% | −0.45% | |
| Retail bought, foreigners and institutions both sold | 3 months | 48.3% | 48.4% | −0.75% | −0.45% |
Just as bad. Stocks retail sold heavily: 43.6% winners after three months, median −2.1%. Indistinguishable from the stocks retail bought heavily (43.6%, −2.0%).
Section 02 is the reason. Retail buys heavily on crash days and sells heavily on spike days. Both signals therefore select "stocks that moved a lot today", and stocks that moved a lot did worse than ordinary stocks over the next three months regardless of direction.
Betting against retail fails not because retail is smart,
but because what retail picks and what retail dumps are the same kind of stock.
05Year by year
| Year | Count | Win rate | Median |
|---|---|---|---|
| 2020 (from July) | 2,300 | 52.3% | +0.30% |
| 2021 | 4,960 | 44.4% | −0.60% |
| 2022 | 4,920 | 42.7% | −0.83% |
| 2023 | 4,900 | 45.6% | −0.43% |
| 2024 | 4,880 | 46.6% | −0.54% |
| 2025 | 4,840 | 51.0% | +0.14% |
| 2026 (to August) | 3,092 | 48.6% | −0.32% |
Only late 2020 and 2025 beat the baseline — both were bull markets where every dip bounced straight back. The other years were at or below it. The top-20 net-sell list looks the same (48.3% in 2020, 41.5% in 2022).
06Caveats
The universe is 749 stocks with a market cap above 100 billion won, reasonable turnover and six years of data. The genuinely small stocks where retail trading concentrates are not here, and they may behave differently.
Net buying is in shares. The top-20-by-value ranking uses shares times the closing price as an approximation.
The "heavy" signals are rare — 4,533 days in six years for the 10% threshold. The direction is clear; year-by-year splits are noisy.
Costs are the 0.22% buy-side deduction only. Delisted stocks are not in the data.
07What I take from it
- Retail buys down days and sells up days: 90% of heavy-buy days were declines
- "Retail net buy" by itself is no information — it matches the baseline
- Stocks retail bought heavily did worse over three months (43.6% vs 48.4%)
- But stocks retail sold heavily did equally badly (43.6%)
- Both are "stocks that moved a lot today". Fading retail does not work
The top-retail-buying list is a fine way to see what crashed today. It is neither a buy signal nor a signal to do the opposite. The data says one thing: a stock that moved a lot yesterday, in either direction, did worse than an ordinary stock over the next three months.
댓글 쓰기