AI Reward Mechanism

Walrus Land introduces a dynamic and performance-driven reward model powered by AI. Instead of distributing static or time-based rewards, the system evaluates player actions and outcomes to determine fair and scalable compensation in $WLAND.

This ensures that players are incentivized not only to participate, but to engage meaningfully and strategically.


Reward Evaluation Criteria

After each PvE battle, the AI Reward Engine analyzes multiple dimensions of performance to calculate an appropriate reward output:

Metric
Description

Difficulty Score

Assesses the strength of the opponent relative to the Walrus's level

Strategy Novelty

Rewards creative or non-repetitive approaches in combat behavior

Consistency Bonus

Factors in historical win/loss trends and behavioral improvement over time

Efficiency Rating

Measures speed and damage taken during combat to evaluate tactical quality


Reward Calculation Formula

The total reward RRR for a single PvE battle is computed as:

R=i=1n[Bi(DiαEiβlog(Ci+2))(1+Ni)]R = \sum_{i=1}^{n} \left[ B_i \cdot \left( \frac{D_i^\alpha \cdot E_i^\beta}{\log(C_i + 2)} \right) \cdot \left(1 + N_i\right) \right]

Where:

  • nnn: Number of combat phases or rounds in the battle

  • BiB_iBi​: Base reward for phase iii (based on enemy tier)

  • DiD_iDi​: Difficulty ratio (enemy power ÷ Walrus power)

  • α\alphaα: Difficulty exponent (amplifies higher challenge tiers)

  • EiE_iEi​: Efficiency score (function of HP remaining × time score)

  • β\betaβ: Efficiency exponent (adjusts reward scaling curve)

  • CiC_iCi​: Strategy reuse count (how often the same tactic was used in recent history)

  • NiN_iNi​: Novelty boost (0 ≤ NNN ≤ 1), measured via vector distance in strategic behavior


Why AI-Based Rewards?

  • Promotes strategic experimentation, not repetitive farming

  • Encourages consistent improvement and adaptation

  • Adds a layer of fairness by rewarding effort and originality

  • Prevents botting and low-effort grinding

This mechanism ensures that earning $WLAND is not only fair — it’s intelligent, skill-aligned, and resistant to abuse.

Last updated