Overview
The Subnet 42 scoring system evaluates miner performance by analyzing telemetry data collected from their TEE (Trusted Execution Environment) workers. This scoring mechanism is designed to reward miners that successfully process X/Twitter data collection tasks and web scraping, while penalizing those with errors or failed operations.
How telemetry data is collected and processed
Telemetry Data Sources: Each registered TEE worker periodically reports telemetry data that includes:
X/Twitter Analytics
Tweet collection statistics
Profile data retrieval metrics
API usage and rate limit tracking
Web Scraping Metrics
Success and failure counts for web scraping operations
Performance tracking across different target sites
Error Monitoring
Authentication failures
Rate limit exceeded events
Network and connectivity issues
Other operational errors
Timing Data
Operation start/end timestamps
Processing duration metrics
Interval between data collections
Snapshot Analysis
System stores multiple telemetry snapshots
Scores calculated from oldest to newest changes
Recent activity and improvements rewarded
TEE Restart Handling
Detects negative delta values
Resets telemetry for fresh start
Ensures non-negative scoring deltas
Scoring Algorithm
Telemetry Data Collection
For each node, the system:
Retrieves telemetry snapshots
Calculates deltas between records
Handles restart scenarios
Normalizes values for scoring
Key Performance Metrics
The scoring system evaluates:
Web Success : Successful web scraping operations
Tweet Collection : Successfully retrieved tweets
Profile Data : Successfully retrieved Twitter profiles
Kurtosis Weighting
A custom kurtosis function weights top performers more heavily in the final scoring calculations.
def apply_kurtosis_custom (
x,
top_percentile = 90,
reward_factor = 0.4,
steepness = 2.0,
center_sensitivity = 0.5,
boost_factor = 0.2
) :
Function Details
Weighting Logic
Applies higher weights to top 90th percentile nodes
Uses configurable curve parameters
Balances rewards for adequate performance
Metric Normalization
Scales values to 0-1 range
Provides minimal non-zero scores
Handles statistical outliers
Score Combination
The final score incorporates:
Web scraping success rate
Tweet collection metrics
Profile retrieval performance
Nodes with balanced performance across metrics receive higher scores, while uneven performance results in moderate scoring.
Validation Process
The system performs these checks:
Minimal scores for low-activity nodes
Score normalization across all nodes
Zero scores for invalid/disconnected nodes
Weight Application
The validated scores are then:
Converted to blockchain weights
Used to determine TAO rewards
Applied through the Bittensor network
Weight Conversion & Updates
Weight Calculation
Scores converted to blockchain weights
TAO rewards based on final weights
Regular update intervals
Update Process
Minimum interval between updates
Up to 3 retry attempts
Score reports sent to miners
Maintain High Uptime
Keep your TEE worker running continuously to avoid service interruptions and restarts
Reduce Errors
Minimize authentication failures, rate limits, and other operational errors
Optimize Success Rates
Focus on achieving consistent success with X and web scraping operations
Balance Metrics
Aim for strong performance across all metrics rather than excelling in just one area
Monitor Performance
Regularly check telemetry data to identify and address potential issues early
Technical Architecture
Core Components
WeightsManager : Handles weight calculations
NodeManager : Manages miner connections
TelemetryStorage : Handles data persistence
ScoringFunctions : Implements scoring logic
Weight Calculation The calculate_weights
method in WeightsManager:
Processes telemetry data
Normalizes metrics
Applies kurtosis weighting
Generates final scores
Scoring Process
Process Telemetry Data
Analyzes changes in miner performance metrics over time using delta-based calculations
Extract & Normalize
Standardizes raw metrics into comparable values across different data types
Apply Weighting
Uses kurtosis weighting to balance consistency with peak performance
Calculate Scores
Combines weighted metrics into comprehensive performance scores
Generate Weights
Converts final scores into network weight allocations for rewards
Conclusion
The Subnet 42 scoring system implements a fair and transparent approach to miner rewards. By combining:
Delta-based performance tracking
Normalized metric analysis
Kurtosis-weighted scoring
The system creates balanced incentives that encourage both consistent reliability and performance excellence in web and social data collection.
Responses are generated using AI and may contain mistakes.