Published on

MEV Bots and the Dark Forest

Authors

I've been building bots to arbitrage profitable function calls on smart contracts. I knew MEV existed-I've seen their addresses in action dozens of times.

What I didn't know is the depth of their operations, and how shallow my understanding of what they exploit really was.

How Bots Compete

It's fascinating how bots compete against each other to claim profitable transactions in a public mempool.

They do this by increasing the fee paid per unit of gas:

  • Reactively - checking what the last bot bid and bidding more
  • Blindly - increasing the bid by a static amount every X seconds

Over time, this drives fees so high the transaction becomes unprofitable. The miner wins, everyone else loses.

So instead, with no formal agreements, bots begin to bid the minimum extra amount the network allows. Leaving it to chance. Reducing gas fees paid to miners.

Game theory in action.

Sophistication Levels

Looking at advanced bots like 0xC0ffeEBABE5D496B2DDE509f9fa189C25cF29671 (c0ffeebabe.eth) reveals a different level of discipline.

These sophisticated MEV hunters don't just use simple frontrunning. They employ complex smart contracts with:

  • Multi-layered obfuscation techniques
  • Transient storage to hide strategy details
  • Dynamic code execution paths
  • Complex gatekeeping mechanisms
  • WETH manipulation capabilities
  • Miner incentivization for better transaction ordering

The level of sophistication is mind-blowing. Examining their bytecode reveals handcrafted assembly specifically built to avoid detection, competition, and disruption.

Flashbots Isn't Enough

While improving my bots, I reached for Flashbots-a promising solution to MEV extraction. But it doesn't prevent it completely.

These same bots monitor not just public mempools, but mined blocks and all transactions within.

This allows bots to re-simulate transactions and uncover patterns over time. I've seen this myself when a profitable transaction from over a day before was copied and resubmitted once found to be profitable again.

The machine learning angle is interesting. Advanced MEV operators use pattern recognition to build predictive models that anticipate profitable opportunities before they even appear in the mempool.

No Easy Defenses

Basic protection mechanisms like token ownership requirements fail because of addresses like 0x3d782C0C69101358a8267Ba116c86726fDF35F91, which take flashloans of tokens to bypass these checks.

This challenges all contracts implementing profitable open transaction calls. The Ethereum is a Dark Forest article by Paradigm outlines this well using CryptoKitties as an example.

Solutions: Fight Fire with Fire

One approach: disallow this behavior at the protocol level. But given humans can't build perfect systems, other exploits will be found and consolidated.

Another: make it simply unprofitable.

These bots are dangerous but blind. Their operators analyze mistakes over time, but that's wasted effort.

One well-crafted script can confuse thousands of bots for days. It's reasonable to build a system where blind bots can't profit in a sea of millions of unprofitable-looking-but-profitable transactions. Only specialized, knowledge-driven bots will win.

Practical Defenses for Contract Designers

Smart contract designers need to implement:

  • Time-locked execution mechanisms - delay between intent and execution
  • Commit-reveal patterns - hide transaction details until execution
  • Strategic honeypot transactions - poison the well for blind bots
  • Unpredictable execution paths - make simulation expensive or unreliable

The MEV landscape is evolving quickly. Understanding these predators is essential to designing systems that maintain economic integrity while discouraging exploitative extraction.

Resources

For more on MEV and baiting bots:

Still building. Still learning. The dark forest is deeper than I thought.

Discuss on Twitter • View on GitHub