Scoring System

The Steward program uses a 4-tier hierarchical scoring system combined with binary eligibility criteria to rank validators for stake delegation. This page provides detailed formulas and explanations for each component.

Overview

The scoring system has two main components:

1. Binary Eligibility Criteria: Pass/fail checks (score of 1.0 or 0.0) that ensure validators meet minimum requirements
2. 4-Tier Hierarchical Ranking: A tiered scoring system that ranks eligible validators with commission being most important, followed by age and performance

Binary Eligibility Criteria

All validators must pass ALL of the following binary checks to be eligible for delegation. Each criterion evaluates to 1.0 (pass) or 0.0 (fail).

MEV Commission Score

mev_commission_score = {
  1.0  if max(mev_commission[t1, t2]) ≤ mev_commission_bps_threshold
  0.0  otherwise
}

where:
  t1 = current_epoch - mev_commission_range
  t2 = current_epoch
  mev_commission_bps_threshold = 1000 (10%)

Purpose: Ensures validators don't charge excessive MEV commissions.

Commission Score

commission_score = {
  1.0  if max(commission[t1, t2]) ≤ commission_threshold
  0.0  otherwise
}

where:
  t1 = current_epoch - commission_range
  t2 = current_epoch
  commission_threshold = 5%

Purpose: Ensures validators don't charge excessive inflation commissions in recent epochs.

Historical Commission Score

historical_commission_score = {
  1.0  if max(historical_commission[t1, t2]) ≤ historical_commission_threshold
  0.0  otherwise
}

where:
  t1 = first_reliable_epoch = 520
  t2 = current_epoch
  historical_commission_threshold = 50%

Purpose: Bans validators who have performed commission manipulation (rug pulls) at any point in their history.

Running Jito Score

running_jito_score = {
  1.0  if any MEV commission exists in [t1, t2]
  0.0  otherwise
}

where:
  t1 = current_epoch - mev_commission_range
  t2 = current_epoch

Purpose: Ensures validators are running the Jito MEV-enabled client.

Delinquency Score

delinquency_score = {
  1.0  if (vote_credits[t] / (total_blocks[t] × TVC_MULTIPLIER)) ≥ scoring_delinquency_threshold_ratio
       for all t in [t1, t2]
  0.0  otherwise
}

where:
  t1 = current_epoch - epoch_credits_range
  t2 = current_epoch - 1
  scoring_delinquency_threshold_ratio = 0.97
  TVC_MULTIPLIER = 1000 (Turbine Vote Credits multiplier)

Purpose: Ensures validators maintain high uptime and vote participation. The ratio measures what fraction of possible votes were cast.

Note: total_blocks is the field in ClusterHistory that tracks how many blocks were created by the cluster in a given epoch. This represents the maximum number of vote credits a validator can earn. Vote credits are synonymous with epoch credits.

Blacklisted Score

blacklisted_score = {
  0.0  if blacklisted in current epoch
  1.0  otherwise
}

Purpose: Allows governance to manually exclude validators for security or policy reasons.

Superminority Score

superminority_score = {
  0.0  if in superminority in current epoch
  1.0  otherwise
}

Purpose: Prevents concentration of stake by excluding validators in the top 33.3% of total network stake, promoting decentralization.

Merkle Root Upload Authority Score

merkle_root_upload_authority_score = {
  1.0  if authority is TipRouter or OldJito
  0.0  otherwise (including Unset, DNE, or other values)
}

Purpose: Ensures validators are using an acceptable Tip Distribution merkle root upload authority for MEV distribution.

Priority Fee Commission Score

priority_fee_commission_score = {
  1.0  if avg_realized_commission[t1, t2] ≤ max_avg_commission OR current_epoch < priority_fee_scoring_start_epoch
  0.0  otherwise
}

where:
  avg_realized_commission = ceiling(sum(realized_commission[t] for valid t in [t1, t2]) / count(valid t))
  realized_commission[t] = ((total_fees[t] - tips[t]) × 10000) / total_fees[t]  (in basis points)
  t1 = current_epoch - priority_fee_commission_range
  t2 = current_epoch

  valid t = epochs where priority_fee_merkle_root_upload_authority is not Unset

Special cases:
  - If tips is None, default to 0 (assumes not distributing priority fees, commission = 10000 bps)
  - If total_fees is None or 0, return 0 bps (no data available yet)
  - If total_fees is None when tips exists, default to u64::MAX (forces commission to max)

Purpose: Ensures validators distribute priority fees to stakers and don't keep excessive amounts. Uses ceiling division to be more strict (rounds up commission).

Priority Fee Merkle Root Upload Authority Score

priority_fee_merkle_root_upload_authority_score = {
  1.0  if authority is TipRouter or OldJito
  0.0  otherwise (including Unset, DNE, or other values)
}

Purpose: Ensures validators are using an acceptable authority for priority fee distribution.

4-Tier Hierarchical Ranking

Validators that pass all binary criteria are ranked using a 4-tier system encoded as a 64-bit value. Higher tiers always dominate lower tiers.

Score Encoding

The score is encoded as a u64 with the following bit layout:

Bits 56-63 (8 bits):  Tier 1 - Inflation Commission (inverted)
Bits 42-55 (14 bits): Tier 2 - MEV Commission (inverted)
Bits 25-41 (17 bits): Tier 3 - Validator Age
Bits 0-24 (25 bits):  Tier 4 - Vote Credits Ratio

This encoding ensures that when comparing scores as u64 values, differences in higher tiers always dominate lower tiers.

Tier 1: Inflation Commission (Highest Priority)

tier1_score = 100 - min(commission_max, 100)

where:
  commission_max = max(commission[t] for t in [t1, t2])
  t1 = current_epoch - commission_range
  t2 = current_epoch

Range: 0-100 (8 bits)

Inverted: Lower commission = higher score

Purpose: Prioritizes validators with lower inflation commissions. A validator with 0% commission scores 100, while 5% commission scores 95.

Tier 2: MEV Commission

tier2_score = 10000 - min(mev_commission_avg, 10000)

where:
  mev_commission_avg = ceiling(sum(mev_commission[t] for valid t in [t1, t2]) / count(valid t))
  t1 = current_epoch - mev_commission_range
  t2 = current_epoch
  valid t = epochs where mev_commission is not None

  If no valid epochs exist, default to 10000 (BASIS_POINTS_MAX)

Range: 0-10000 basis points (14 bits)

Inverted: Lower commission = higher score

Purpose: Among validators with equal inflation commission, prefers those with lower MEV commissions. Uses ceiling division to be more strict.

Tier 3: Validator Age

tier3_score = min(validator_age, 131071)

where:
  validator_age = count of epochs with non-zero vote credits

Range: 0-131,071 epochs (17 bits, ~716 years max) Direct: Older validators score higher Purpose: Among validators equal on commissions, rewards longevity and operational track record.

Tier 4: Vote Credits Ratio

tier4_score = min(vote_credits_avg_scaled, 33554431)

where:
  vote_credits_avg_scaled = vote_credits_ratio × VOTE_CREDITS_RATIO_MAX
  vote_credits_ratio = (sum(vote_credits[t] for t in [t1, t2]) / count[t1, t2]) /
                       (sum(total_blocks[t] for t in [t1, t2]) / count[t1, t2]) /
                       TVC_MULTIPLIER

  t1 = current_epoch - epoch_credits_range
  t2 = current_epoch - 1
  VOTE_CREDITS_RATIO_MAX = 10,000,000 (for precision)
  TVC_MULTIPLIER = 1000 (Turbine Vote Credits multiplier)

Range: 0-33,554,431 (25 bits)

Direct: Higher performance = higher score

Purpose: Among validators equal on all other tiers, rewards better performance. This is a relative measure of validator reliability and vote participation, not an exact APY measure.

Final Score Calculation

The final score combines the 4-tier raw score with all binary eligibility factors:

// Step 1: Encode 4-tier raw score
raw_score = (tier1_score << 56) | (tier2_score << 42) | (tier3_score << 25) | tier4_score

// Step 2: Apply binary filters
final_score = raw_score
  × mev_commission_score
  × commission_score
  × historical_commission_score
  × blacklisted_score
  × superminority_score
  × delinquency_score
  × running_jito_score
  × merkle_root_upload_authority_score
  × priority_fee_commission_score
  × priority_fee_merkle_root_upload_authority_score

Result: If any binary criterion fails (= 0), the final score becomes 0, making the validator ineligible. Otherwise, the final score equals the raw_score, which can be compared directly as a u64 to rank validators hierarchically.

Example Comparison

Here's how two validators would be ranked:

Validator A: 1% commission, 5% MEV commission, 100 epochs old, 0.95 vote ratio Validator B: 2% commission, 3% MEV commission, 200 epochs old, 0.98 vote ratio

Despite Validator B having better MEV commission, age, and performance, Validator A ranks higher because Tier 1 (inflation commission) dominates all other factors. The 1% vs 2% difference in inflation commission overrides all other advantages.

Implementation Reference

See the full implementation in score.rs in the StakeNet repository.