How We Score Crypto Projects: Inside the CryptoVerdict Research Engine
Ask ten analysts to rate the same crypto project and you'll get ten different answers — usually shaped less by the underlying fundamentals than by who's paying, who's holding, and which Twitter thread set the narrative that week. That's the core problem CryptoVerdict was built to solve.
This post is a walk-through of how our research engine actually works: what we look at, how we weight it, why we cite every claim, and how to read a verdict without being misled. If you've ever wondered what separates a thoughtful AI report from a generic chatbot answer, this is for you.
Why most crypto research is biased
Crypto research has a structural problem: most of the people producing it are also positioned to profit from it. Newsletters and Telegram callers buy first and write later. "Independent" researchers accept paid coverage from foundations. Influencers are rewarded by engagement, not accuracy, and engagement loves hype.
Even well-meaning analysts run into the same limitation: there's simply too much information, scattered across too many sources, changing too fast for any one person to track honestly. So they take shortcuts — they anchor on price, on social momentum, on which fund just announced a position. None of that is fundamentals. All of it is reflexive.
We built CryptoVerdict on a different premise: that the underlying evidence is almost always public, and that the bottleneck isn't access, it's synthesis. If we could automate the synthesis — rigorously, transparently, and without an axe to grind — we could give every investor the same starting point a professional analyst would have.
Why evidence matters more than hype
Hype is a leading indicator of nothing except more hype. Evidence is what survives when the narrative changes.
A real research process answers concrete questions: What does the protocol actually do? Who built it, and what's their track record? How is the token distributed, and to whom does value accrue? Has the code been audited, and by whom? Is anyone using it, on-chain, in volumes that aren't being faked?
Every one of those questions has an answer somewhere — in documentation, in GitHub commits, in audit reports, in on-chain data, in governance forum posts. The job of a research engine is to find those answers, weigh them, and present them in a form a human can act on. The job is not to predict the price.
That distinction matters. We don't claim to know whether a token will go up. We claim to tell you, with citations, what's actually there.
Our ten research pillars
Every CryptoVerdict report scores a project across the same ten pillars. The framework is intentionally fixed so that two different tokens can be compared on the same axes — something that's almost impossible with free-form research.
- Technology. What problem does the protocol solve, and how? Is the architecture credible? Does the codebase show active, competent development?
- Team. Who is building this? Are they doxxed, anonymous, or somewhere in between? What have they shipped before?
- Tokenomics. Supply schedule, distribution, unlock cliffs, value accrual. Does holding the token actually capture the protocol's success, or is it cosmetic?
- Security. Audit history, audit quality, incident history, bug bounty programs, multisig and key management practices.
- Adoption. On-chain activity, TVL, transactions, unique users — and whether those numbers are organic or incentivized.
- Community. Engaged users vs. paid engagement farms. Quality of developer ecosystem and third-party integrations.
- Governance. Who actually controls the protocol? How are decisions made, and how meaningfully decentralized is the process?
- Roadmap. Is there a credible plan, with shipped milestones backing it up? Or is it perpetual "coming soon"?
- Competition. Who else is solving this problem, and why would a user pick this one?
- Catalysts. Concrete upcoming events — mainnet launches, upgrades, token unlocks, partnerships — that could materially change the picture.
Each pillar gets a 0–10 score grounded in cited evidence. The pillars roll up into a single overall verdict, but the value is in the breakdown: a project can be brilliant on technology and broken on tokenomics, and you deserve to see both.
How confidence scores work
Every score on a CryptoVerdict report carries an implicit confidence signal: how much underlying evidence the engine was able to find and verify.
A protocol with multiple audits, a public team, years of GitHub history, and detailed documentation produces a high-confidence report — the pillars are graded against a lot of source material, and disagreement between sources is rare. A six-week-old token with an anonymous team, no audit, and a one-page website produces a low-confidence report — not because we refused to grade it, but because the evidence is thin.
We surface this directly. Sections cite the sources they're based on, and when evidence is missing or contradictory, the report says so rather than guessing. A "we don't know" is a feature, not a bug. The worst outcome in research isn't an unknown — it's a confident answer that turns out to be wrong.
When you read a verdict, the question to ask isn't just "what's the score?" — it's "what's the score based on?" The citations are there for exactly that reason.
Public Reports vs. Full AI Research
CryptoVerdict serves research in two modes, and the difference is important.
Public Reports are reports that have already been generated for a project and are available to the whole community. They're free to browse (within a daily limit), instantly accessible, and great for discovery — if you're scanning the market or evaluating something for the first time, a public report is almost always the right starting point. The trade-off is timeliness: a public report reflects the evidence available at the moment it was generated, which could be hours or weeks ago.
Full AI Research re-runs the full pipeline against the latest available evidence, on demand, for one credit. You get the most current scores, the most current citations, and a private copy saved to your library that you can re-run, export, and compare against later. This is the premium experience and the right call when you're about to actually act on a token — entering a position, sizing up, evaluating a catalyst.
The rule of thumb: public reports for discovery, full research for decisions.
Why you should always verify before investing
Even the best research engine — ours included — is a tool, not a decision. A CryptoVerdict report exists to compress the work of due diligence, not to replace your judgment.
Markets move on things research can't always see: liquidity dynamics, counterparty risk, sudden regulatory shifts, behavioral effects that have nothing to do with fundamentals. A high score tells you the underlying project is credible. It does not tell you that the trade is good, that the timing is right, or that the price is fair.
Use our reports the way a professional investor uses a research desk: as a rigorous starting point, not as the final word. Read the citations. Click through to the original sources. Form your own view. Size your positions accordingly. And remember that nothing in any of our reports is financial advice — it's structured information designed to make you a better, more skeptical reader of the market.
That's what evidence-backed research is for. It doesn't tell you what to do. It tells you what's true, as far as anyone can verify it. The decision is, and always should be, yours.
