1) After analyzing 82 million trades on Upbit XRP/KRW, I found something very interesting... A $5 billion one directional selling pipeline running 24/7 for almost a year Read along 🧵

2) It started with yesterday's price action. -57M XRP in CVD over 17 hours. It looked insane. So I ran forensic queries - bot fingerprinting, iceberg detection, wash trade checks The selling was real. Algorithmic. 61% of trades fired within 10ms. Single bot running 17 hours straight with one 33 second pause

3) But then I zoomed out. And that's where it gets wild. -57M isn't an anomaly. Upbit XRP/KRW has been net negative every single month for 10 months Apr: -165M Jul: -197M Oct: -382M Jan: -370M Total: 3.3 BILLION XRP in net selling. ~$5B

4) Only 1 week out of 46 was positive. One. And it's not "the market" - Binance XRP/USDT carries 2-5x less sell pressure on the same coin (shocker) In June, Binance was net positive while Upbit bled -218M The hourly correlation between the two venues is only 0.37. Upbit's flow is largely its own thing

5) Here's where it gets interesting. From April to September, Upbit $XRP was trading 3-6% BELOW Binance. A reverse Kimchi discount The sellers were accepting 6% worse fills than available on global markets, for many months They don't care about the price. They need KRW, are mandated to use Upbit, and/or are Korean holders taking profit...

7) I bucketed every day by what XRP did on Binance globally: Crash days (<-5%): -46M avg CVD, 1.49x sell/buy Down days: -22M avg CVD Flat days: -6M avg CVD Green days (2-5%): +4M avg CVD, 0.94x sell/buy Moon days (>5%): +8M avg CVD, 0.93x sell/buy

8) Read that again. On moon days, Korean retail becomes a NET BUYER. They're accumulating On crash days, sell intensity is 8x heavier The systematic seller + retail panic amplify each other Korean retail buys every rip. The pipeline sells into all of it

9) The top 15 heaviest sell days on Upbit XRP/KRW: Feb 5, 2026: -147M (1M trades) Jul 23: -87M Oct 10: -75M (the day the premium flipped) Jan 31: -53M Aug 14: -51M The worst day was Oct 7 and it had a 2.08x sell/buy ratio. For every 1 $XRP bought, 2 were sold. On a venue with millions of retail users

10) The bots have been identical for 10 months: 57-60% of all trades fire within 10ms Same round number sizes: 10, 50, 100, 500, 1000 XRP 24/7 operation, no weekday/weekend distinction There is no time of day where buying outweighs selling in aggregate This isn't a trader. It's infrastructure

11) So who's on the OTHER side? The buy side tells a story 28% of buy trades are tiny fractional sizes: 2.535, 3.679, 2.681 XRP. This is consistent with KRW-denominated orders (buy 10,000 won of XRP). The sell side? Round numbers. 10, 100, 1000 XRP. It looks mechanical. One side looks like retail. The other looks like a machine.

12) Ten million fractional buy orders over 10 months Compared to the sell side running mechanical round number clips Two completely different profiles trading against each other on the same venue

13) Korean investors can't easily access global exchanges (capital controls). So Upbit trades at a premium to Binance. The sellers collect that 2-3% spread on top of spot price The premium = the cost of being in a walled garden

14) To put the scale in perspective: 3.3 billion XRP = 5.4% of XRP's entire circulating supply Net sold through a single trading pair, on a single exchange, in 10 months ("Net sold" also = shifted hands)

15) So who has enough XRP to sell 300-400M per month for a year straight, doesn't care about 6% discounts, runs identical algo infrastructure 24/7 and needs KRW specifically or is in some walled garden and can only use Upbit? AND who are they selling to? Who's been on the other side of that trade? It could be 1 entity, 50 entities or 10k people I'll let you speculate

16) This analysis used tick trade data I collected from Upbit and Binance 82M Upbit trades + 444M Binance trades, tick level precision Thanks for reading (like & follow appreciated)