Every sandwich on BNB Chain, documented. Block by block.
An open effort by BNB Chain builders, validators, searchers and researchers. We document toxic MEV from public on-chain data, attribute it, and publish recommendations. Every number is public, every rule is versioned.
Every computed day, on the record.
One bar per UTC day, computed once the day has closed.
View as table
ConfirmedThe back-run reverses the front-run within 1% at a profit, and the two legs share an actor.
ProbableProfitable, with one of those signals. Counted, and labelled as such.
Pattern onlyThe shape alone. Not counted anywhere on this site.
Every block, checked the same way.
Rules anyone can read, queries anyone can re-run.
Detect
Every block, every DEX pool. We look for the same-pool bracket: an attacker's front-run, a victim's swap, then the attacker's back-run.
Attribute
From the block's own structure: which builder built it, which validator sealed it, which address took the profit.
Publish
Every figure comes from a public query on Dune. Rules are versioned, and each sandwich records the version that caught it.
Recommend
Findings come with recommendations to builders, validators, the Alliance and wallets. We don't enforce them; the record shows whether they were followed.
Who builds the blocks that carry sandwiches.
Raw counts follow market share. The share of a builder's own blocks that carry a sandwich is the figure that compares.
| Builder | Blocks built | Share of blocks | Sandwiches | Blocks with a sandwich |
|---|
A block is attributed to the builder that paid for it: each builder's payment sits inside the block itself. Any builder named here can reply, and its reply will be shown next to its figures.
Full scorecard, with validators, on Dune →Whose users get sandwiched.
Every swap is attributed to the app that built it: a wallet, a trading app, an aggregator or a DEX's own site. The app chooses the RPC that broadcasts the swap and the partners that route it.
| App | Swaps | Sandwiched | Sandwiched per 10,000 swaps |
|---|
The ten named apps with the most sandwiched swaps, sorted by rate. Apps are named only from a public source; integrator ids appear as written on-chain.
Every app on Dune →The Alliance sets the standard. We keep the record.
Two efforts, one goal. According to BNB Chain, the Good Will Alliance cut sandwich attacks by about 95%. We document what is left, in public.
The commitment
- Ethical standards for block building on BNB Chain
- A public list of MEV-protected builders
- Validators accept blocks only from builders on the list
The record
- Every block documented, every day, from public data
- Named rates per builder, validator and app, with a right of reply
- Recommendations to the Alliance, builders and validators, who decide what to act on
Recommended practice, role by role.
We have no power to enforce anything. We document what happens on-chain and recommend what should change; builders, validators, the Alliance and wallets decide what to do with it.
Builders
- Filter sandwich and front-running bundles out of every block.
- Tag blocks and declare bid addresses, so attribution is exact.
- Publish a response when a finding concerns you.
Validators
- Accept bids only from builders with a clean record.
- When you build a block yourself, build it clean.
- Publish the list of builders you accept.
Searchers
- Don't submit sandwich or front-running bundles.
- Arbitrage, liquidations and back-runs are out of scope.
- Respond when an address you control is documented.
Wallets & dApps
- Route swaps through MEV-protected endpoints by default.
- Default to tight, sensible slippage.
- Tell users when a trade was sandwiched.
Members.
Members have joined the effort. Each one contributes to the analysis as far as its technical means allow, and to the fight against sandwich attacks and toxic MEV. Being a member changes nothing in how anyone is documented.
The member list is published at launch.
Don't take our word for it.
Every number on this site traces back to raw chain data through public queries. If a figure is wrong, you can show us exactly where.
Our commitmentEverything in the open. No side taken. Only facts.
Versioned rules
Rule changes are numbered and dated. Old findings keep the version that produced them.
Read the methodology →Disputes, in public
Think a finding is a false positive? File it. The case and the decision are published.
Dispute channel opens at launchHelp keep the record.
Share data, review the method or challenge a finding. Builders, validators, searchers, wallets, dApps and researchers are all welcome.