A small cluster of 12 Solana bot addresses saw a 2.97x higher rate of positive WSOL balance changes when routing through the proprietary AMM HummingFi, according to a study accepted at ASE 2026. The researchers examined 200 addresses — the top 100 from Trojan and the top 100 from SolanaMevBot — tracking transactions from Oct 1 to Nov 1, 2025. The finding suggests that venue choice can sharply separate outcomes even on a single blockchain.
What the study found
Transactions invoking HummingFi had a positive WSOL balance change 62.3% of the time. Other transactions from the same group of addresses saw a 21.01% rate. The 12-address cluster routed heavily through Jupiter and also invoked proprietary AMMs including HummingFi. The profit measure was the change in WSOL token balance before and after a transaction — not a fully netted account of strategy returns.
The limits of the data
The study shows a correlation within the cluster, not proof that HummingFi access caused the gap. The researchers published a replication package on Zenodo. They also note the study does not connect HummingFi or the sampled bots to harms like sandwich attacks on retail users. So the advantage could come from the cluster's strategy, not the AMM itself.
A separate cluster on Pump.fun
A different 102-address cluster concentrated 80.9% of its activity on the Pump.fun ecosystem. That group's behavior was distinct from the HummingFi cluster, reinforcing how venue choice sharply separated the sampled bots' activity — even on one blockchain.
For ordinary DEX users, the finding suggests that participants on one chain can encounter different opportunities when their routing, infrastructure, or venue set differs. The study doesn't say whether the HummingFi cluster's edge will persist, but it raises a question: is the advantage in the AMM's design or in the bots' own tactics? The replication package is out there for anyone to dig into.
