📖 Methodology
Normalize, weight, gate, tier: asset-class-aware. Scores run 0 to 100: each feature is a percentile against the asset class's reference basket, averaged per domain, weighted with that class's weights, then capped by any gate that trips. Every token page shows the working behind each number.
Domain weights (default / DeFi)
| class | fund.fundamental | tokn.tokenomics | chainonchain | socialsocial | devdev |
|---|---|---|---|---|---|
| defi | 30% | 25% | 20% | 10% | 15% |
| l1 | 18% | 20% | 25% | 12% | 25% |
| monetary | n/a | 35% | 35% | 12% | 18% |
| meme | n/a | 20% | 35% | 45% | n/a |
| stablecoin | n/a | n/a | 65% | 35% | n/a |
Tiers
Hurdle
10Y treasury 4.46% (as of 2026-06-18). A token paying real yield below it is flagged.
Asset classes
Gate: hard disqualifiers
Gates marked inactive are wired but cannot fire on free data: contract verification is True-or-unknown (Sourcify can't prove a negative), and team-anonymity facts have no keyless source.
Metric glossary
- P/Flower is better
Market cap divided by annualized fees: how expensive the token is per dollar of user activity.
- P/Slower is better
Market cap divided by the annualized revenue the protocol keeps.
- MC / TVLlower is better
Market cap divided by total value locked. Under 1 looks cheap, over 5 looks rich.
- Real yieldhigher is better
Holder yield funded by real revenue, not token emissions.
- FDV / MCAPlower is better
Fully diluted value divided by circulating value. High means a large dilution overhang.
- Unlock overhanglower is better
Share of max supply still locked behind vesting, which is future sell pressure.
- Unlock / volumelower is better
Dollar value of the next unlock divided by daily volume. Can the market absorb it?
- Floathigher is better
Circulating supply divided by total supply. Low float plus high FDV is the classic trap.
- Inflationlower is better
Annual new supply divided by circulating supply.
- Token sinkhigher is better
Share of revenue routed back to holders (buyback / burn / staking).
- Top-10 holderslower is better
Supply held by the 10 largest wallets. High = whale / dump risk.
- Active-address growthhigher is better
Daily active addresses: last-30-day mean vs first-30-day mean of a 90-day window.
- Exchange reserveslower is better
Trend in exchange-held supply. Declining = accumulation.
- Social trendhigher is better
Trend in social volume and mentions (Santiment).
- Sentimenthigher is better
CoinGecko community up-vote share (keyless, coarse).
- Watchlistshigher is better
How many CoinGecko users track the token: a broad, free attention and adoption signal.
- Dev activityhigher is better
Average developer-activity events per day over a 90-day window, ranked against class peers. A sustained level: a busy or quiet month on its own is not a signal.
- Dev-activity trendhigher is better
Change in developer activity across the window (informational, not scored).
- Days since last pushlower is better
How recently the team shipped code (gate input).
🧠 Break your thesis
- User retention: what happens if users leave? Is liquidity sticky or mercenary?
- Supply shifts: what happens at the next unlock? Are early investors incentivized to dump?
- Regulation: does it depend on regulatory ambiguity? What if it is deemed a security?
- Bear-market survival: can it survive a severe drawdown? Alts fall harder than BTC.
- The final question: if everything you just worried about were true, would you still own it?
Frequently asked
What is a DYOR score?
A 0 to 100 composite for one crypto token. Each feature (for example price-to-fees or holder concentration) is ranked as a percentile against a fixed reference basket of same-class tokens, features are averaged per domain (fundamentals, tokenomics, on-chain, social, developers), domains are weighted with class-specific weights, and any hard disqualifier that trips caps or zeroes the result.
What do the tiers mean?
The tier is a fixed mapping of the final score: A (high conviction) is 80 and above, B (qualified) 60 to 79.9, C (watchlist) 40 to 59.9, D (avoid) below 40. Tiers are a research shorthand, not a buy or sell recommendation.
Why isn't Bitcoin judged on revenue like a DeFi protocol?
Tokens are classified first (DeFi protocol, L1 / platform, monetary, memecoin, stablecoin) and each class scores only the dimensions that matter for it. A monetary asset has no fundamentals domain, so missing protocol revenue cannot hurt it; a DeFi app with no measurable fees, revenue or TVL has its fundamentals domain floored instead of skipped.
What are the gates?
Hard disqualifiers checked after the composite: extreme FDV/MCAP (above 10x) caps the score at 40; no audit on record at DefiLlama caps a DeFi protocol at 50; a dead token (no push in 180 days on any GitHub account found for it with no Santiment developer activity to contradict it, or near-zero daily volume) zeroes the score. Missing data never trips a gate.
Where does the data come from?
Free, open sources only: CoinGecko (market data, supply, community votes, watchlists), DefiLlama (fees, revenue, holders revenue, TVL, audits; parent protocols aggregate their versions), Santiment (daily active addresses and developer activity over a 90-day window), GitHub (most recent push), Ethplorer (top-10 holder shares), Sourcify (contract verification). Every token page lists each source's status and links to its page.
Why do scores move between weeks?
Because the token's own data moved: a changed TVL, fee run-rate, usage trend or supply figure. The reference baskets that scores are ranked against change only when they are deliberately rebuilt, percentiles are interpolated and treat near-identical values as ties, and a source that fails on a given run has its last stored values carried forward and labelled stale, so outages and provider rounding do not move a score.
See it applied: every scored token or analyze one live.