Upbit Listings

Upbit tracks how prices on Binance and Bybit react to new listing announcements on Upbit, Korea's largest exchange. The pattern is consistent enough to archive: price spikes hard within minutes of the announcement, then fades.

Every announcement becomes a row, with the windows around it measured automatically. Times are shown in KST, Upbit's own timezone, because that is how the announcements are published.

The page is not a screener — nothing here is a daily pattern. It is an archive of one repeating event, plus what the archive says about trading it.

The market a listing opens matters more than anything else on this page. Announcements that include the KRW market bring Korean retail flow and are the real spike — historically a median +34% peak, with 90% of them moving at least 10%. Announcements limited to USDT/BTC are much weaker, at a median +11%.

That is why the summary counts KRW listings only, and why the KRW listings only toggle exists — the rest are noise for this purpose.

The summary also carries the archive's verdict: of the KRW listings that actually moved, every one gave ground back from its peak, by a typical −25%, and 62% peaked inside the announcement minute — which is why chasing an alert usually means buying the high. The lookup beside it sizes that expectation by how hard the coin spiked: the bigger the spike, the more comes back, from −16% on moves under 20% to −42% on moves over 100%.

Below the table, the page derives a single rule from the archive and shows its working. The headline answers the obvious question first — why not just long it? Because the jump belongs to whoever is already there: longing the announcement instantly is profitable, and the identical trade two seconds later is not. That is a bot on a direct feed, not someone reading an alert. What is left for everyone else is the plateau afterwards, and that is a short.

The rule itself is computed, not fixed — the page picks it from a grid of combinations, requiring a minimum number of trades, no liquidation at 2×, and profitability in both halves of the history, then ranks by the weaker half. So the numbers you see are whatever currently survives that test, not a number someone chose.

How much leverage is a table, not a recommendation. For a fixed margin per trade it shows the fixed-stake and compounded results side by side with two columns that matter more: dies if it runs and liquidated. A row that earns more but shows liquidations is not a better row.

Does the filter hold up is the robustness check — the same target measured at every filter level, with the trade count next to each. It is there so you can see whether the edge survives when the gate moves, or whether it only exists at one convenient setting.