Reading a sybil report a project publishes
Every airdrop season, a handful of projects publish a sybil report after the token generation event. It’s usually a blog post or a page on the docs site, sometimes a PDF from a data vendor the project hired, and it lists how many addresses got excluded or clawed back and, in the better ones, why. Most farmers skim the headline number and move on. That’s a mistake. The methodology section is the only place a project tells you, in writing, what its detection actually looked for, and reading it closely is worth more than any “how to avoid sybil detection” thread on Twitter.
This isn’t a guide to beating a specific report’s heuristics. It’s how to read one so you understand what it measured, what it didn’t, and what that means for how you think about wallet hygiene generally.
What a sybil report actually is
A sybil report is a project’s after-the-fact accounting of which addresses it decided weren’t real independent users. It’s published for a few reasons that aren’t mutually exclusive: to justify allocation cuts to a community that’s watching total supply and asking where the excluded tokens went, to deter farming ahead of the next distribution round, and increasingly because the project used a third-party clustering vendor and is passing along that vendor’s summary as a transparency gesture.
The report format varies a lot. Some are a single paragraph with a percentage. Others run pages long with charts, example cluster diagrams (addresses anonymized or hashed), and an appeal process for wallets that think they were flagged incorrectly. The length and specificity of the report is itself a signal, more on that below.
Why the methodology section is the part worth reading
The headline number, “we excluded 4.2% of claiming addresses,” tells you almost nothing on its own. What matters is the paragraph or two describing how those addresses were identified. Real sybil detection on public chains is built from a fairly small set of signals, and a report that names its signals specifically is giving you far more useful information than one that just says “advanced heuristics and machine learning.”
The signals that show up again and again in published methodologies:
Funding source clustering. Wallets that all received their initial gas or seed capital from the same upstream address, especially in a tight time window, get grouped. This is the single most common thing named in reports because it’s cheap to compute and hard to fully avoid if capital has to originate somewhere.
Timing correlation. Transactions across “different” wallets that land in the same block, or a few seconds apart, night after night, read as coordinated rather than independent. A report might phrase this as “addresses exhibiting synchronized interaction patterns.”
Behavioral near-duplication. Identical sequences of contract calls, the same order of interactions across a protocol’s features, the same gas price down to a few gwei, or the same slippage settings on every trade. Individually weak, but a strong signal when stacked with the others.
CEX withdrawal linkage. Some reports mention working with an exchange or a chain analytics firm to trace withdrawal batches, where a single exchange transaction fans out into dozens of addresses that all go on to behave identically. This is one of the harder signals to see coming, because it happens off the protocol entirely.
Interaction shallowness. A count of how many distinct actions, contracts, or days an address touched. Reports sometimes phrase exclusion criteria around a minimum activity bar rather than clustering at all, which is a different mechanism worth telling apart from the others.
Notice what’s usually absent: device fingerprinting, IP address, or browser signals almost never appear in a public methodology, even when the project’s anti-sybil vendor is clearly using them. That’s not because the signal isn’t in play, it’s because disclosing exactly how a fingerprinting vendor works would help people build around it, so it stays undisclosed. If a report is unusually vague about “additional off-chain signals” right where it explains its on-chain clustering, that’s worth registering. It usually means there’s a browser or infrastructure-level layer under the on-chain layer, even if it’s never spelled out.
Reading the numbers without over-trusting them
A typical report leads with something like “18,000 addresses excluded, representing 3.4% of eligible wallets.” Treat the shape of that number as more informative than its precision. Ask three questions.
First, is the denominator eligible claimants or total addresses that ever touched the protocol? Those are very different bases, and reports aren’t always clear which one they’re using.
Second, does exclusion mean zero allocation or a reduced one? Some projects don’t fully zero out flagged clusters, they discount them, which changes what the clawback number is actually counting.
Third, is there an appeals window, and did the project publish results after appeals closed? A report published the same day as the claim, before any appeal has run, is a first draft. A report updated weeks later with a revised number is telling you the first pass had false positives worth correcting, which is useful context on how noisy the detection was.
None of this means the report is wrong. It means a single percentage without that context is closer to a marketing figure than a technical one, and it’s worth reading past it before drawing conclusions about how aggressive or accurate the project’s detection was.
What this tells you about tooling, not just wallets
The reason these reports are worth reading isn’t to reverse-engineer a specific project’s filter. Methodologies get updated every cycle and a technique that worked last quarter tells you little about the next one. What’s useful is the pattern across many reports over time, because it tells you which categories of infrastructure choices are structurally correlated and which aren’t.
If report after report names shared funding sources and synchronized timing as its top two signals, that’s a strong argument for treating funding paths and scheduling as the highest-leverage things to think about when you’re running several wallets for legitimate, independent activity, more so than any single browser fingerprint setting. If a report separately calls out identical RPC endpoints or default wallet client fingerprints as part of its off-chain layer, that tells you something about why RPC provider choice and browser configuration show up as a recurring theme in this space at all, they’re not cosmetic, they’re one more axis a clustering pipeline can pull on.
This is also why anti-detect browser and RPC provider reviews matter as a category, and why we test them here rather than take a vendor’s word for it. A tool’s marketing claims about “unique fingerprints” or “residential-grade IPs” are unrelated to whether it actually changes the signals a real sybil report says it looks at. Reading published reports is how you calibrate what’s worth testing for in the first place.
The limits of what gets published
No public sybil report is a complete picture of detection. Projects don’t publish their full pipeline because doing so would make it easier to route around, and most reports are written for a token holder audience worried about dilution, not for a technical audience trying to understand chain analysis. Expect selective disclosure, rounded numbers, and methodology descriptions written in plain language that smooth over a lot of engineering underneath.
There’s also real disagreement in the space about false positive rates. Genuine multi-wallet users, people who split funds across addresses for reasons that have nothing to do with farming, get caught in these nets regularly, and a report’s stated exclusion count includes an unknown number of them. Reports rarely quantify this, and when they do it’s usually only after an appeals process surfaces enough complaints to be worth mentioning.
What to actually do with this
Read the methodology paragraph before the headline number. Note which signal categories a project names, on-chain and off-chain, and track that across reports over a few cycles rather than reacting to any single one. Use what you learn to think generally about funding structure, timing, and tooling choices for wallets you’re running as separate, real activity, not as a checklist to defeat one vendor’s current filter. And don’t treat any of this as a guarantee. A published report is a project’s account of what it caught last time, not a rulebook for what any future project will check, and nothing here is financial advice about whether a given token or airdrop is worth your time.
If you want more of this kind of breakdown, along with tested reviews of the anti-detect browsers, RPC providers, and trackers that actually come up when people are thinking through wallet ops, you’ll find it on the Airdrop Farming home page.
Get new guides and videos first — join the Telegram channel.