Two-thirds of Avalanche's trading comes from two clusters of bots
Dashboards show billions of dollars changing hands on Avalanche's exchanges each month. We wanted to know whose dollars they are.
- 65%
- 40% to 67%
- 8 in 10
- 32%
- $347,000
- 8%
01 · The tradingAvalanche DEX volume, and who sends it
Avalanche looks like a busy place to trade. Its exchanges handled $3.6 billion of trades in September. A year earlier, in their busiest month, they handled more than four times as much.
Avalanche is a blockchain, and most of its apps run on one part of it, the C-Chain. Its decentralized exchanges, or DEXs, let anyone swap one token, a coin or other asset kept on the chain, for another. Each swap trades against a pool, a store of two tokens that other people have put in, and each one is written to the chain.
Bitquery reads every transaction on Avalanche's main chain, so we rebuilt each swap and traced who sent it. For the six biggest DEXs, our totals land within 1.2% of DefiLlama's, a site that tracks DEX volume. Then we sorted the senders.
Most of the money comes from very few of them. In late September, Avalanche set a price floor, a higher minimum fee for every transaction. In the 8 days after it, 65% of the dollars traded came from two groups of wallets, the accounts that send trades on a chain. One address paid for each group, which makes each group what on-chain analysts call a cluster. Both groups behave like bots, programs that trade on their own, and both trade in step with the price of AVAX, Avalanche's own coin, on Binance, the largest crypto exchange.
The clusters are not new. The newer one arrived in August 2025. A month later, in the busiest month the chain's DEXs have had, it made 36% of the trading. Since then the two together have made at least 40% of every month's DEX volume.
02 · Two clustersTwo addresses behind 366 wallets
A bot can spread its trades across many wallets, and each wallet needs a little AVAX to pay its gas, the fee every transaction pays. So we looked at where the busiest traders got their AVAX. Two addresses stood out.
One funder is a plain wallet that has sent AVAX to a couple of hundred wallets since early 2024, more than 21,500 times. The other is a contract, a program that lives on the chain, which paid its wallets between August and October 2025. No wallet got AVAX from both.
Each cluster sends its trades through its own contracts. One cluster sent 99.6% of its transactions to a single contract, so counting the trades by contract gives the same 65% as counting them by wallet. None of the four contracts has published its code, and the chain does not say who runs the clusters.
Everyone else shared what was left. Other bots made 27% of the dollars. The rest of the wallets, which pass none of our bot tests, made 8%, about $12 million per day.
03 · What they doThey trade like arbitrage bots, a step behind Binance
Arbitrage bots buy a coin where it is cheap and sell it where it is dear, and the Avalanche half of each trade leaves a trail on the chain. On the morning of 29 September, AVAX climbed 5% on Binance in an hour. In that same hour the clusters bought $3.9 million of WAVAX, Avalanche's coin in token form, and sold $2.1 million, buying hardest in the minutes the price jumped.
The clusters trade WAVAX more than anything else. Most of their dollars, 83%, went into WAVAX against USDC, a dollar stablecoin. Most of the rest was wrapped ether and bitcoin against WAVAX.
Every transaction they traded in after the floor touched one pool, so none of them traded one Avalanche pool against another in a single step. Both clusters also sell about as much as they buy. In those 8 days, one cluster's buying and selling of WAVAX came out level to within $100,000, and the other bought 8% more than it sold.
Across all 8 days, we lined up their trades with AVAX's price on Binance, minute by minute. In 8 of every 10 minutes they traded, each cluster bought on Avalanche when AVAX rose on Binance and sold when it fell. Their buying also tracks Binance's move from the minute before, but Binance's next minute shows almost no link to what the clusters did. Binance moves first, and the clusters follow.
That is the pattern of arbitrage bots. Traders call this kind CEX-DEX arbitrage, with a centralized exchange, or CEX, such as Binance on one side and a DEX on the other. When AVAX jumps on Binance, an Avalanche pool sells it too cheaply for a moment, and an arbitrage trading bot buys there until the two prices meet. A market maker, a firm that quotes both buy and sell prices, can leave a similar trail if it keeps its holdings level. The Binance side of each trade, if there is one, is not on the chain, so we can see the pattern but not the whole trade.
04 · What it costsA third of Avalanche's gas fees
Every transaction on Avalanche pays its gas in AVAX, and the chain burns it. In late September the validators, the computers that run the network, raised the lowest price a transaction can pay. Bots as a whole sent less than half as many transactions after that. These two clusters kept going.
On a typical day after the floor, the clusters paid about a third of all the gas fees the chain burned, the same share as before. Their gas bill rose with everyone's, and one cluster's more than tripled. 48% of the other cluster's transactions failed, and a failed transaction still pays.
Gas is the smaller cost. Each swap also pays a pool fee. Most of the clusters' trades go through Pharaoh, an exchange on Avalanche, and in its busiest pool they paid about $347,000 of pool fees in 8 days, about seven times their gas. That pool is set to send the whole fee to Pharaoh itself rather than to the people who supply its tokens, and the chain does not show who ends up with it. The clusters do not supply that pool either. None of its deposits and withdrawals in those days came from them.
| What both clusters paid, 27 Sep to 4 Oct | Amount |
|---|---|
| Gas fees, typical day | 580 AVAX, about $6,400 |
| Share of all gas fees, typical day | 33% before, 32% after |
| One cluster's gas fees, typical day | 84 AVAX before, 297 AVAX after |
| Failed transactions, the other cluster | 48% |
| Pool fees in Pharaoh's busiest pool | about $347,000 in 8 days, all set to go to Pharaoh |
What the clusters earn is not on the chain. If they sell on an exchange what they buy on Avalanche, that half of each trade happens where we cannot see it.
05 · After the floorFewer swaps, bigger trades
In the 8 days after the floor, swaps per day fell by more than half, and the dollars traded per day rose by a third. The average swap tripled in size.
All of the lost swaps came from outside the clusters. The clusters made about as many swaps per day as before, while everyone else made 64% fewer. The extra dollars came from the clusters too. Everyone else traded about the same dollars per day as before, even with AVAX 19% higher, while the dollars each cluster traded per day rose by 50% and 72%. A cluster's trade is large, so gas is a small part of it.
| Avalanche's exchanges, per day | 13 to 25 Sep → 27 Sep to 4 Oct |
|---|---|
| Swaps | 295,648 → 126,398 |
| Dollars traded | $117.7 million → $156.2 million |
| Average swap | $413 → $1,253 |
| Swaps by both clusters | 30,457 → 30,798 |
| Two clusters' share of dollars | 54% → 65% |
| AVAX price, average | $9.27 → $10.99 |
Robinhood Chain went the same way. When its fee price rose 25 times, most of the extra demand came from eight addresses running trading software.
06 · One exchangePharaoh, the exchange that took over
Pharaoh is one of several DEXs on Avalanche. In April it handled a quarter of the trading, and LFJ, formerly Trader Joe and then the largest, handled 39%. By September Pharaoh had 81%. Two of its pools, both for WAVAX against USDC, handled 62% of all the chain's DEX volume that month.
The clusters put 83% and 89% of their trading through Pharaoh. The chain shows where they trade, and it does not show why. You can follow trades like these on any exchange with a DEX trades API.
07 · What it measuresWhat Avalanche's DEX volume measures
Most of the money traded on Avalanche comes from two clusters that follow Binance's price. That is real trading, with real tokens changing hands, but it says little about how many people trade there.
The wallets that do not trade like bots made 8% of the dollars in those 8 days. Avalanche is not alone. On Optimism, bots have made 86% of exchange trading since July, and on Arc, a new chain, two wallets made half of one day's volume.
If you want to know how many people trade on a chain, count the wallets and see who sends the dollars.
08 · How we measured itMethod
We rebuilt every swap on Avalanche's C-Chain from January 2025 to 4 October 2026, for Pharaoh, LFJ, Blackhole, Uniswap, Pangolin and smaller DEXs, and counted only pools made by known exchanges. One unknown contract claimed $2.28 billion of trades in a single day in May 2025 while no WAVAX moved, and pools like it are left out. Each swap counts once, on its dollar side, with stablecoins at one dollar and AVAX, bitcoin and ether at Binance's daily price.
A cluster is every wallet one address ever paid AVAX to. A bot, by our loosest test, sent 50 or more transactions on a typical active day, or 1,000 on any day. For timing, we compared the clusters' buying and selling each minute with AVAX's move on Binance. The correlation, a score of how closely two series move together, is 0.58 and 0.61 for the same minute and about zero for Binance's next minute. Gas fees come from every transaction the clusters sent, compared on a typical day, the median, because one sender's bidding on 24 September burned 6,278 AVAX, more than six times a typical day.
The DEX trades and contract events docs describe the fields we used. Our look at Avalanche's real-world assets uses the same data.
09 · The recordAddresses behind this story
FAQ
Who trades on Avalanche DEXs?
Mostly bots. In the 8 days after the price floor in late September 2026, two clusters of bot wallets made two-thirds of Avalanche DEX volume, and wallets that do not trade like bots made 8%.
Are the Avalanche clusters arbitrage bots?
They trade like crypto arbitrage bots. They trade one pool at a time, buy and sell in similar amounts, and in 8 of every 10 minutes they trade, they buy when AVAX rises on Binance and sell when it falls. What they do on Binance, if anything, is not visible on the chain.
Who runs the clusters?
The chain does not say. One cluster's funder is a plain wallet and the other is a contract, and none of the clusters' contracts has published its code.
How much do bots pay in Avalanche gas fees?
On a typical day after the price floor, the two clusters paid 580 AVAX of gas, about a third of all the gas fees Avalanche burned.
What is Pharaoh exchange?
Pharaoh is a DEX on Avalanche. It handled 81% of all the chain's DEX volume in September 2026, up from a quarter in April.
Did the price floor cut Avalanche's DEX volume?
Not in dollars. After the floor, swaps per day fell by more than half, while the dollars traded per day rose by a third and AVAX's price rose 19%. Both clusters kept trading and paid about a third of all gas fees.
Arbitrage is our reading of the pattern: the clusters trade one pool at a time, buy and sell in similar amounts, and follow Binance's price within the minute. A market maker that keeps its holdings level could leave a similar trail. What they do on any exchange is not visible, so we do not know what they earn.
A cluster is every wallet one address ever paid AVAX to. Counting the same trades by the contracts they go through gives the same shares, which tells us the wallets work together. It does not tell us who runs them.
Our bot tests count transactions per day. A careful person can pass for a bot and a slow bot for a person, so the split between other bots and everyone else is a range, not a line.
The after-floor window is 8 days, 27 September to 4 October 2026. AVAX's price was 19% higher than in the two weeks before, which lifts every dollar total.
We count trades on exchanges that write swaps to the chain. Exchanges that settle trades another way, about 5% of Avalanche's trading by DefiLlama's count, are left out. For LFJ's newest pools our total is higher than DefiLlama's; it matches the tokens that moved in and out of the pools.
Fees use the price each transaction offered, which after the September upgrade runs about 0.3% above the price it paid. Fee comparisons use median days because of the 24 September spike. Pool fees are measured in one pool, Pharaoh's busiest, which the public node shows sending the whole fee to Pharaoh; who receives it after that is not visible.
This article is provided for informational and educational purposes only and reflects analysis of publicly available on-chain data as of the dates indicated. It does not constitute legal, financial, compliance, or investment advice, and nothing in it is a recommendation to buy, sell, or hold any token.
The findings describe swaps and transactions observed in Avalanche C-Chain data over the period stated, compared with public Binance prices and DefiLlama totals. Avalanche addresses are pseudonymous, and an unlabelled address is described by its behaviour and attributed to nobody.
References to Pharaoh, LFJ, Blackhole, Uniswap, Pangolin, Binance, DefiLlama or any other named party describe on-chain records and public data, and are not statements about any entity's conduct or intentions. Calling a pattern arbitrage describes trades, not any person or company.
Nothing herein should be relied upon as a definitive determination of fact. Readers should conduct their own independent verification before taking any action. The authors and publisher accept no liability for any loss or damage arising from reliance on this material. All trademarks and company names are the property of their respective owners.
Reported by Gaurav Agarwal for Bitquery Research, with AI tools; every figure was checked against the raw data.
Trace any trader on Avalanche
The swaps, wallets and fees above come from Bitquery's Avalanche C-Chain data, which holds every transaction since 2020. The same data is open through our API, so you can rebuild any exchange's trading, follow a wallet back to whoever paid for it, or measure how much of a pool's volume comes from bots.