Independent Market Assessment
Never self-reported by data providers. Computed by Atlas across five independent dimensions every 5 seconds. No single source can influence its own score.
CONSENSUS SCORE
Prices You Can Verify Trust You Can Measure
Consensus Score is Atlas Oracle's real-time market confidence signal that measures the reliability of oracle price feeds before protocols make critical on-chain decisions.
A Price You Can Reason About
What Is Consensus Score?
Consensus Score is Atlas Oracle's real-time market confidence signal. Alongside the price, it measures exchange agreement, liquidity conditions, and market quality, so protocols know not just what the price is, but how much to trust it.
Consensus Score vs Traditional Oracle Price Feeds
Traditional oracle price feeds primarily answer one question: “What is the current price?” Consensus Score goes one step further by helping protocols evaluate the quality and reliability of that price before making critical on-chain decisions.
| Dimension | Atlas Oracle | Standard Oracle |
|---|---|---|
| Output | Price + Consensus Score + 6-level health status + reason code | Price only |
| Question answered | How trustworthy is this price, and why? | What is the price right now? |
| Assessment basis | System-computed independently across 5 dimensions | Publisher-reported or single-dimension |
| Manipulation detection | Cluster analysis + volume concentration + drift + split-market flag | Range widens, cause unknown |
| Actionability | 6-level status maps directly to risk-control actions | Low. Tells you what, not what to do |
Transparent Reason Codes
When the score degrades, the reason is explicit: insufficient sources, price split, stale data, or volume anomaly. Respond to the cause, not the symptom.
Programmable Risk Signals
The 6-level health status maps to concrete on-chain actions. Smart contracts branch on status in a single conditional. No custom parsing required.
First-Party Market Coverage
375+ CEX and 570+ DEX connections via the CoinMarketCap platform, one of the broadest market-data networks in the industry.
Five Dimensions
How Is Consensus Score Calculated?
Consensus Score evaluates five independent dimensions of market quality to measure how reliable an oracle price feed is in real time. Instead of relying on a fixed formula, Atlas dynamically adjusts the weighting of each dimension based on current market conditions and source availability.
Data Sufficiency
Are there enough active, independent sources to support a reliable price? Fewer sources means higher weight on this dimension.
Price Agreement
Do sources agree on the price? High dispersion, or a detected split market, downgrades this dimension sharply.
Time Consistency
Are prices arriving in real-time? Sources with stale or cached data are tracked and penalized.
Volume Health
Is there real trading activity behind the price? A price that looks stable but has no volume is flagged.
Source Quality
Are sources consistently reliable? Historical reliability feeds into the current score.
Safety Mechanisms
Veto power
Any single dimension below 0.10 caps the total score at 0.30. One failing dimension cannot be masked by others.
EMA smoothing
Sharp score drops respond fast; recovery is gradual. Prevents gaming the system with a brief “clean” window.
Split-market detection
Automatic cluster analysis identifies when sources form two distinct price camps, the clearest signal of manipulation.
Health Status
Six States. One Conditional.
Every feed publishes a health status alongside the score. The status maps directly to a risk-control action, so your contract can respond automatically without human judgment.
Score ≥ 0.85
Multi-source agreement. All dimensions appearing normal. Fully trustworthy for normal protocol operations.
→ Normal operation
0.65 – 0.85
Slight quality degradation. Price is likely still accurate but reliability has declined.
→ Increase monitoring
0.40 – 0.65
Significant issues in one or more dimensions. Raise collateral, reduce leverage, or pause new positions.
→ Tighten risk controls
Score < 0.40
Price may be unreliable. Atlas publishes the last trusted value and halts advancement.
→ Pause / circuit breaker
Special state
Cluster detection has identified two distinct price camps, the primary signature of active manipulation.
→ Investigate immediately
Insufficient data
Not enough source data to compute a reliable score, such as during a feed's cold-start window.
→ Safe default: no output
Status transitions include a ±0.05 hysteresis band to prevent rapid oscillation. Upgrading to HEALTHY requires ≥ 0.90, and downgrading requires a score below 0.80.
Validation
Tested Against Market Stress
ConsensusScore has been backtested against documented historical market anomalies across five distinct attack and failure patterns. Each shows what the score would have reported and what protocols could have done automatically.
Scenario 01 · Manipulation
Cross-venue price split (DEX inflation)
SPLIT_MARKET
A DEX price was driven to ~7× the prevailing CEX price. Centralized sources kept reporting the correct price; decentralized sources reported the manipulated one.
→ Pause usage / trigger circuit breaker
CONSENSUSSCORE OUTPUT
Score: 0.10
Status: SPLIT_MARKET
Reason: Two distinct price camps detected (CEX vs DEX)
Scenario 02 · Manipulation
Low-liquidity price inflation (+2000%)
WARNING
A token with limited exchange coverage was pumped >2000% on a single low-volume venue to inflate collateral. Contained to one source with minimal real trading.
→ Tighten risk controls
CONSENSUSSCORE OUTPUT
Score: ~0.45
Status: WARNING
Reason: VolumeHealth low (manipulated venue excluded, <1% VWAP)
Scenario 03 · Data integrity
Single major-exchange outlier (777% de-peg)
DEGRADED
One large exchange reported a >700% de-peg for a liquid staking token. All other sources stayed normal. A standard oracle would have weighted this heavily.
→ Increase monitoring
CONSENSUSSCORE OUTPUT
Score: ~0.75
Status: DEGRADED
Reason: SourceQuality issue on one venue (outlier excluded, ±10% median)
Scenario 04 · Exchange failure
Exchange API failure: stale data
CRITICAL
An exchange went offline but its APIs kept returning outdated prices. Other oracles propagated these stale prices without flagging, since the value looked reasonable.
→ Pause / activate circuit breaker
CONSENSUSSCORE OUTPUT
Score: <0.40
Status: CRITICAL
Reason: Stale data, price + volume unchanged (GRACE → DEGRADING → STALE)
Scenario 05 · Market dispersion
Pool vulnerability: 15–30% price gaps
SPLIT_MARKET
Liquidity-pool vulnerabilities caused 15–30% price gaps across exchanges. The spread widened gradually before escalating into full market fragmentation.
→ Monitor at DEGRADED; investigate at SPLIT_MARKET
CONSENSUSSCORE OUTPUT
Score: falls as gap widens
Status: SPLIT_MARKET
Reason: 15% gap lowers Price Agreement; 30% gap → 2 camps
Protocol Use Cases
Your Risk System, Automated
ConsensusScore feeds directly into automated risk controls. The 6-level status eliminates custom thresholds or off-chain monitoring for common scenarios.
Lending Protocols
HEALTHY
Normal liquidations and collateral reads proceed.
WARNING
Raise collateral ratio to require more backing per loan.
CRITICAL
Pause liquidations. Price may be unreliable.
Derivatives & Perps
HEALTHY
Normal trading and mark-price settlement.
WARNING
Reduce maximum leverage on affected pairs.
CRITICAL
Pause new position opens on an unreliable mark.
Cross-chain Bridges
HEALTHY
Normal transfer processing for all amounts.
WARNING
Queue large transfers for manual review.
CRITICAL
Pause that asset bridge. Don't value it unreliably.
375+
CEX connections via CoinMarketCap platform
570+
DEX connections available on demand
1s
Score update frequency, every batch cycle
8
Checkpoints before a price reaches your contract
Get Started
Build Like the Price Can't Betray You.
Integrate ConsensusScore into your protocol's risk system. Testnet setup in 1–3 days. Full integration in 2–4 weeks with dedicated technical support.