---
title: Is the bundle selling into your buy? Counter-trading on pump.fun, measured
seo_title: pump.fun counter-trading: is the bundle selling into your buy?
description: A bundler told Halibut his bundle sells 0.8 SOL into every 1 SOL you buy. We checked 630,275 pump.fun coins since BOOST launched. 782 did it, 0.9% of those graduated, and the real problem in new pairs is something else.
date: 2026-10-27
category: Bundles
facts: [counter_trade]
allow: [10, 25, 30, 70, 140, 250, 80, 0.8, "1 SOL", "0.8 SOL", "3 SOL", "$10k", "$70k", "$140k", "$250k", "$25k", "30 SOL", "$10M", "80%", "10%", "20%", "30%", "$60k", "$6M"]
---
## Quick answer

Some bundles do sell into every buy, and it's rare. Since pump.fun launched BOOST, 630,275 coins got 20 or more outside buys in their first 6 hours. In 782 of them (0.12%) the launch bundle answered most of those buys with a sell inside 2 slots, at a median 0.69 SOL back per SOL bought. Those coins graduated 0.9% of the time, against 15.96% for everything else. Counter-trading is real, it farms the smallest new pairs, and it is not why your feed is full of "runners" that die in an hour. That's something else, and it got about 62 times worse when BOOST went live.

## What Halibut's bundler said

[Halibut](https://x.com/HalibutCrypto/status/2114912988222382248) talked to a bundler who went from $10k days to $70k days to a $140k day while Halibut ate a $250k loss. The pitch, via a reflink to a launch tool: buy 25 aged, CEX-funded wallets for 3 SOL, bundle 10%, 20%, 30%, even 80% of the supply, then turn on "counter trading": every time someone buys 1 SOL of your coin, your bundle automatically sells 0.8 SOL. The chart still goes up, just slower, and nobody notices. Repeat all day, make 30 SOL on $25k toppers.

The reply under the article asked the right question: "do we have good tools to check if the coin is bundled?" Halibut: "good question, idk". So we checked.

## Is the bundle selling into your buy?

A bot that sells into your buy leaves a fingerprint: bundle sells that land in the slots right after outside buys, and not in the slots before them. Normal sellers dump whenever. A bot answering buys only shows up after.

So for every coin with at least 20 outside buys in its first 6 hours, we took the bundle (the creator, every wallet that bought on the curve in the first 10 seconds, and every wallet that sold the coin without ever buying it: aged wallets that got the tokens by transfer) and asked, for each outside buy, did a bundle wallet sell within 2 slots after it? Then the same question for the 2 slots before it. A coin is flagged when at least 50% of outside buys get answered, 3 times the before-rate, at a replication ratio between 0.2 and 1.5.

| | coins checked | flagged | share | flagged graduated | everyone else graduated |
|---|---:|---:|---:|---:|---:|
| since BOOST (2026-07-20 to 2026-10-24) | 630,275 | 782 | 0.12% | 0.9% | 15.96% |
| same, answer window 10 slots | 630,275 | 1,236 | 0.2% | | |
| before BOOST (25 sampled days, 2025-03-21 to 2026-07-10) | 187,594 | 260 | 0.14% | | |

The tell is the before/after split, and it points the wrong way for the bundler story. Across all coins the median share of outside buys followed by a bundle sell is 12.4% after the buy and 13.7% before it. In 61 of 62 two-day windows since BOOST, sells landed before buys at least as often as after. Ordinary two-way trading, no bot.

Where counter-trading does run, it pays a little: the 782 flagged bundles netted a median 0.87 SOL each, 3,979.1 SOL between them. The flagged count runs 2 to 30 per two-day window and hasn't trended up since we started looking in March 2025 (0 to 40 then). It's an old trick on a new reflink.

## What one looks like, checked on chain

The biggest one we found: [mro1rHzwDiTGTbzC2627N3tuaaVxLtW2fzNsDxUpump](https://pumpdata.fun/check/mro1rHzwDiTGTbzC2627N3tuaaVxLtW2fzNsDxUpump), launched 2026-08-10. The bundle took 79.3% of the supply at launch. Of 1,368 outside buys, 63% were answered within 2 slots (5% before), and over 4,669.93 SOL of outside buying the bundle netted 1,211.59 SOL. It graduated anyway.

Three of its pairs, each a 90-100 SOL buy answered 1 or 2 slots later by a different bundle wallet:

| outside buy | bundle sell | slots later | sold back per SOL bought |
|---:|---:|---:|---:|
| 100.07 SOL | 90.54 SOL | 2 | 0.91 |
| 97.59 SOL | 85.00 SOL | 1 | 0.87 |
| 93.99 SOL | 103.49 SOL | 2 | 1.10 |

We pulled the first pair from the chain: the pool took in 100.27 SOL in slot 438486672 and paid out 90.36 SOL in slot 438486674. That's the "0.8 SOL into every 1 SOL" machine, live. On one coin out of hundreds of thousands.

## Are pump.fun coins bundled?

Yes, duh. In October 2026, 20.26% of launches had their first-10-second buyers holding 10% or more of the supply, 9.84% had 30% or more, and 5.07% had more than half. The 50%+ share peaked at 9.73% in August 2026, up from 2.35% in March 2025.

And bundles work. Since BOOST, launches whose first-10-second buyers held 50%+ graduated 21.43% of the time, against 1.38% for launches under 10%: 15.5 times as often. That's the bundler's business model. They don't need to sell into you on the curve; they need you to buy the graduation, and then they sell. Our [bundles post](https://pumpdata.fun/research/pump-fun-bundles) shows what happens the day after.

## Has pump.fun printed a $10M runner this month?

Halibut's teaser said the market hasn't printed a 10m runner in a month. It has, if you count a runner as a coin a crowd actually held at $10M. We counted a launch as a runner when its hourly median market cap (trades of 0.2 SOL or more) held a level for 12 hours with 50+ traders each hour, inside 7 days of launch, on $5M+ of volume. Launches from October 2026 that cleared $10M that way:

| coin | held for 12 h | 7-day volume | daily high on GeckoTerminal |
|---|---:|---:|---:|
| [Agency](https://pumpdata.fun/check/7VertkgF9KLhxxJXHX6uaWuoYZTP9LdGj2bWmVXVpump) | $17.9M | $68.4M | $32.1M |
| [AGENTCAT](https://pumpdata.fun/check/CZJfrAmMs5Cz4H22PWzfqTJu5FkP7Q2Vvx1oPVL9pump) | $23.9M | $14.2M | $43.0M |
| [swordinu](https://pumpdata.fun/check/2BVfJ4AHMvHdKtEZNHaBr48dQzTfZvYkjaaxbM6bpump) | $18.3M | $15.2M | $25.1M |
| [CRAWLER](https://pumpdata.fun/check/CQCDTp8rsFV3PhaPeYjcCW5LdGTxi6jFx7eC3UXnpump) | $19.1M | $14.6M | $32.6M |
| [VEIL](https://pumpdata.fun/check/HgTGdc29D5bDMrDvsCLnorxz899VyDj7ZN7FPa6vpump) | $11.5M | $14.8M | $19.4M |
| [STREAM](https://pumpdata.fun/check/9a98oQHtxfqmr4EyzsP6wfw69uYAK3JbUeVEuNW1pump) | $23.6M | $17.6M | $41.0M |
| [MINTER](https://pumpdata.fun/check/3NxfWMHmfQg8NJFW5AG1ti2xACdczWz24e9SKyk1pump) | $13.3M | $13.7M | $22.4M |
| [startupSI](https://pumpdata.fun/check/EKosQw4oJFbXmeJ4qRnbH1quGFRTc3Eok6nwnFKDpump) | $21.0M | $19.7M | $37.0M |

That's 8. Since BOOST, 20 launches held $10M, 31 held $5M and 126 held $1M, out of 3,477,695. None held $50M.

## So what's actually farming new pairs?

Fake runners. Since BOOST, 7,569 coins printed a $10M+ market cap for an hour and never held it: 25 to 352 every two days, about 130.5 per window. In the 25 days we sampled before BOOST it was 53 in total, about 2.1 per window. That's the brain bleed Halibut describes: a feed full of tickers that look like they ran.

Some of those prints are real money doing weird things. One wallet put 5,001 SOL into a single buy on a pool minutes after migration, printing a $865M market cap for an hour, and early holders sold into it. We checked it on chain; it happened. Most of the rest are thin pools moved by a few trades. Either way, the hourly high on a chart is not a runner.

## How do you check a coin?

Paste the contract address into [pumpdata.fun/check](https://pumpdata.fun/check). You get the bundle (how many wallets bought in the first 10 seconds and how much of the supply they still hold), the insiders' share, whether they sold together in one slot, the deployer's record (launches, graduations, best run), and how launches that looked the same at this age went over their next day. Counter-trading isn't a field yet: at 0.12% of coins it would read zero almost every time. If that number moves, we'll add it.

Edge vs edge, like Halibut says. Ours is reading every trade before you buy.

## Frequently asked questions

### What is counter-trading on pump.fun?

A launch bundle that automatically sells part of every outside buy back into it, so the chart still rises, just slower, while the bundle cashes out. Launch tools sell it as a "replication factor": sell 0.8 SOL into every 1 SOL bought.

### How common is counter-trading on pump.fun?

Rare. Since BOOST launched, 782 of 630,275 coins with 20 or more outside buys (0.12%) showed it, at a median 0.69 SOL sold back per SOL bought. In 61 of 62 two-day windows, bundle sells landed before outside buys at least as often as after them.

### Do counter-traded coins graduate?

Almost never: 0.9% of flagged coins graduated, against 15.96% of the rest. The bundles netted a median 0.87 SOL each.

### Has pump.fun had a $10M runner recently?

Yes. 8 coins launched in October 2026 held $10M or more for 12 hours with 50+ traders an hour, among them CRAWLER, STREAM and AGENTCAT. Since BOOST, 20 launches did, while 7,569 printed $10M for an hour and never held it.

### How do I check if a pump.fun coin is bundled?

Paste its contract address into pumpdata.fun/check: it shows how many wallets bought in the first 10 seconds, how much of the supply they still hold, whether insiders sold together, and the deployer's record.