Scoring protocol 2026.07-alpha.1
How ProtoRadar turns evidence into a signal
ProtoScore evaluates project evidence. Signal Strength measures unusual market activity. Evidence Confidence shows how much required proof is available. Missing critical evidence withholds the score.
Read a result in this order
- 1Check Evidence Confidence
- 2Read ProtoScore
- 3Inspect Signal Strength and gates
- 4Confirm the classification
ProtoRadar is not affiliated with, endorsed by, or sponsored by Robinhood or Flap. Platform names describe observable lifecycle context only; they never raise a score by association.
Mental model
Three measures, three questions
Each number has one job. Momentum cannot improve project evidence, and broad coverage cannot make unfavorable evidence favorable.
Project evidence
ProtoScore
0–100- Answers
- How strongly the available evidence supports the project’s quality and credibility.
- Publication rule
- Withheld if any critical pillar has less than 35% evidence coverage.
- Does not mean
- A safety probability or return forecast.
Market activity
Signal Strength
0–100- Answers
- Whether current activity is unusual and broadly distributed relative to the token’s own baseline.
- Publication rule
- A numeric score requires all four inputs; qualification also requires every integrity gate to pass.
- Does not mean
- Project quality or likely price direction.
Coverage
Evidence Confidence
0–100- Answers
- How much required evidence is present, fresh, and sourced at the right level.
- Publication rule
- Reported from 0–100. Missing, stale, or overly broad sources reduce it.
- Does not mean
- That well-covered evidence is favorable.
Decision path
Evidence moves through four explicit stages
The classifier does not jump from attention to a verdict. It records what happened, checks how much can be verified, calculates separate measures, then applies hard gates.
- 01
Observe
Collect contract, market, wallet, product, and optional social evidence from named sources.
- 02
Measure coverage
Check source presence, freshness, and granularity. Unknown evidence remains unknown.
- 03
Calculate separately
Produce ProtoScore, Signal Strength, and Evidence Confidence without letting one substitute for another.
- 04
Apply gates
Return withheld, monitoring, provisional early, qualified, or rejected with the reason preserved.
Lifecycle context
Public markers such as a Flap launch, bonding-curve progression, DEX graduation, or attention from Robinhood Chain ecosystem accounts can describe where a token sits in its lifecycle. They do not replace contract, wallet, or market evidence and are not endorsements.
Project evidence
ProtoScore weights what can be verified
Five pillars define the target model. Unobserved pillars stay unknown, and critical coverage controls whether a public score can appear.
- 25%Transaction integrityTests whether volume reflects varied economic activity rather than repeated, circular, or tightly paired trades.
- 20%Wallet qualityMeasures wallet history, funding provenance, concentration, and effective participation after clustering.
- 20%Contract and market riskReviews contract identity, executable trading paths, liquidity, concentration, and observable controls.
- 20%Product evidenceLooks for utility claims backed by inspectable product, technical, code, or delivery evidence.
- 15%Social authenticityEvaluates account continuity, audience quality, and attention from established ecosystem participants.
Publication gate
Critical coverage controls the public score
ProtoScore is withheld if transaction integrity, wallet quality, or contract and market risk has less than 35% coverage. Partial evidence pulls the result toward the conservative prior of 35; it never promotes an unknown pillar.
Calculation detailsFormula and conservative prior
Each pillar has score s, fixed weight w, and observed coverage c. The observed result uses only evidence the model can see.
Observed Score equals the sum of weight times coverage times score, divided by the sum of weight times coverage.
Evidence Confidence is 100 × Σ(w × c). Normalized confidence q blends the observed result with the conservative prior.
ProtoScore equals confidence times Observed Score, plus one minus confidence times the conservative prior of 35.
Market activity
Signal Strength looks past raw volume
The score asks whether activity is unusual and broadly supported. A qualified pulse also requires every market-integrity gate to be observed and passing.
- 30%Spike ratioRecent qualified volume compared with the token’s own trailing baseline.
- 25%Robust z-scoreHow unusual the move is against a median-based history that resists one-off outliers.
- 25%Effective tradersDistinct economic participants after concentration and related-wallet adjustment.
- 20%Qualified volume to market capQualified volume divided by measured circulating market cap; FDV-only records do not qualify.
Required pulse gatesAll five must be observed and pass
Sudden volume is the trigger, not the conclusion. A numeric Signal Strength requires all four weighted inputs, while qualification also requires every rule below.
- 01Spike ratio is at least 3× and robust z-score is at least 3.
- 02Qualified volume is at least the greater of $750 or 5% of market cap.
- 03At least 8 effective traders remain after cluster adjustment and the top three clusters control no more than 35%.
- 04Fresh-wallet buy volume is at most 20%, fresh buyers at most 30%, and common-funder share at most 8%.
- 05Wash-risk is low, buy share is between 35% and 85%, and real sellability has been observed.
Classification policy
Qualified has a reproducible threshold set
Classification follows gate order. An observed integrity failure rejects the candidate; an unknown required gate withholds it; complete passing evidence can then qualify.
- 01 Rejected
- An observed wallet, concentration, wash-risk, balance, or sellability gate failed.
- 02 Withheld
- Required evidence or an integrity gate is unknown.
- 03 Qualified
- Every qualified threshold passes. This is an investigation status, not an endorsement.
- 04 Provisional early
- A measured early-band candidate clears every provisional threshold and hard gate.
- 05 Monitoring
- Evidence is accumulating or the candidate does not meet the higher classifications.
Exact qualified checklistNo market-cap band applies
No market-cap band applies. Qualified remains a research status, not a safety certification, financial recommendation, or return forecast.
- ≥ 70 ProtoScore
- ≥ 70% Evidence Confidence
- ≥ 35% Each critical pillar
- ≥ 75 Transaction integrity
- ≥ 70 Wallet quality
- ≥ 65 Contract and market risk
- ≥ 60 Product evidence
- ≥ 55 Social authenticity
- ≥ 99.99% Signal Strength input coverage
- Pass All signal gates
- ≥ 65 Signal Strength when the classification depends on a volume claim
Early-candidate policy
Live observation starts at $10K
The $10K–$12K range is the intended earliest review window, not a candidate status or buy signal. The same evidence and integrity rules still apply.
Below $10K
Outside the live feed
The current live feed excludes pools below $10K. ProtoRadar does not build or retain a baseline for them.
$10K–$12K
Earliest research focus
ProtoRadar starts paying closer attention here. This band does not automatically raise or qualify a token.
$10K–$250K
Provisional eligibility
Measured circulating market cap must fall inside this range, and every provisional threshold must pass.
Exact provisional checklistMeasured market cap required
- ≥ 75 ObservedScore
- ≥ 60 ProtoScore
- ≥ 55% Evidence Confidence
- ≥ 50% Average critical coverage
- ≥ 35% Each critical pillar
- ≥ 65 Signal Strength
- Pass All hard gates observed
Market cap and FDV are different. An FDV-only record is labeled as a proxy and cannot satisfy measured market-cap eligibility.
Evidence checks
Risk flags remain evidence-bounded
Behavioral patterns can raise risk or discount volume, but ProtoRadar does not claim manipulation, ownership, or endorsement without direct evidence.
Suspected wash activity
Looks for rapid reversals, repeated sizes or timing, circular counterparties, little net position change, and volume growth without matching trader or liquidity growth. A match raises risk or discounts volume; it does not prove manipulation.
Fresh-wallet concentration
Looks at wallet age, prior activity, shared funders, funding timing, and cluster-like trading. New wallets are not inherently suspicious; concentration plus shared behavior raises risk.
Current definition: A wallet is fresh when it is under 7 days old and has fewer than 10 prior meaningful transactions.
Limits
What the model cannot prove
The methodology makes uncertainty visible, but source gaps, coordinated actors, and offchain behavior can still evade the checks.
- 01Indexers can lag, omit pools, or report incomplete supply and transaction data.
- 02Related actors can use old wallets, multiple funders, private relays, or offchain coordination that resists clustering.
- 03A working contract, public repository, audit, or verified account does not guarantee honest operators or durable utility.
- 04Observed sellability and liquidity are snapshots; later contract, balance, or market changes can invalidate them.