ApacheProject-v0

Category: Validated Case Study (TR-3) Agents: 8-40 (phase-dependent) Difficulty: Advanced Source: coopetition_gym/envs/collective_action_envs.py

Overview

ApacheProject-v0 is a validated case study based on the Apache HTTP Server project (1995-2023). This environment reproduces the 52/60 validation score from TR-3, modeling contributor dynamics across four project phases.

The environment tests whether agents can learn phase-appropriate contribution strategies that match historically observed patterns.

Validation Score

TR-3 Validation: 52/60 (86.7%)

This environment’s parameters are calibrated against real Apache project data, making it a gold standard for TR-3 research.

MARL Classification

Property Value
Game Type N-player Markov Game (phase-structured)
Cooperation Structure Open source contribution dynamics
Observability Full
Communication Implicit
Agent Symmetry Symmetric within phase
Reward Structure Team production with phase-specific loyalty
Action Space Continuous: $A_i = [0, 50]$
Horizon Finite, T = 60 steps (monthly resolution)
Canonical Comparison Open source dynamics; Lerner & Tirole (2002)

Project Phases

Phase Configuration

Phase Years Core Team Size Loyalty Multiplier Expected Effort
Emergence 1995-1999 8 1.00 35.0
Growth 2000-2005 15 0.85 30.0
Maturity 2006-2015 40 0.70 22.0
Evolution 2016-2023 35 0.60 18.0

Phase Dynamics

Emergence (1995-1999):

Growth (2000-2005):

Maturity (2006-2015):

Evolution (2016-2023):

Environment Specification

Basic Usage

import coopetition_gym
import numpy as np

# Create environment for specific phase
env = coopetition_gym.make("ApacheProject-v0", phase="maturity")

obs, info = env.reset(seed=42)

print(f"Phase: {info['phase']}")
print(f"Years: {info['phase_years']}")
print(f"Core team size: {env.n_agents}")
print(f"Expected effort: {info['expected_effort']:.1f}")

# Run episode
for step in range(60):
    # Contribute at expected level
    expected = info['expected_effort']
    actions = np.full(env.n_agents, expected)
    obs, rewards, terminated, truncated, info = env.step(actions)

    if terminated or truncated: break

print(f"Validation accuracy: {info['validation_accuracy']:.1%}")

Parameters

Parameter Default Description
phase “maturity” Project phase to simulate
max_steps 60 Monthly timesteps
render_mode None Rendering mode

Phase Selection

# Emergence phase (small, high-loyalty team)
env = coopetition_gym.make("ApacheProject-v0", phase="emergence")

# Growth phase (expanding team)
env = coopetition_gym.make("ApacheProject-v0", phase="growth")

# Maturity phase (large stable team)
env = coopetition_gym.make("ApacheProject-v0", phase="maturity")

# Evolution phase (maintenance mode)
env = coopetition_gym.make("ApacheProject-v0", phase="evolution")

Spaces

Observation Space

Type: Box Dtype: float32

Dimension varies by phase (based on team size).

Action Space

Type: Box Shape: (n_agents,) where n_agents is phase-specific Dtype: float32 Range: [0.0, 50.0] for each agent

Uniaxial Treatment: This environment uses the single-dimension action space characteristic of Coopetition-Gym v1.x. Contributor competition for recognition and influence emerges through phase-dependent loyalty dynamics rather than explicit competitive actions.

Metrics and Info

The info dictionary contains:

Key Type Description
phase str Current phase name
phase_years str Historical year range
phase_loyalty float Maximum loyalty for phase
expected_effort float Historically calibrated effort
effort_deviation float Actual vs expected deviation
validation_accuracy float 1 - deviation (higher is better)
team_output float Production Q(a)
mean_loyalty float Average loyalty score

Key Dynamics

Phase-Specific Loyalty

Loyalty is capped by the phase multiplier:

loyalty_scores = np.clip(loyalty_scores, 0.0, phase_loyalty_multiplier)

This captures the empirical observation that project maturity affects contributor commitment levels.

Validation Metric

The environment tracks how closely agent behavior matches historical patterns:

deviation = |mean_effort - expected_effort| / expected_effort
validation_accuracy = max(0, 1 - deviation)

High validation accuracy indicates realistic simulation.

Research Applications

ApacheProject-v0 is suitable for studying:

Empirical Validation

TR-3 Validation Framework (60 points)

Category Points Description
Free-riding baseline 10 Matches theoretical equilibrium
Loyalty dynamics 15 Evolution patterns match data
Effort differentiation 15 High vs low loyalty difference
Team size effects 10 Scaling behavior validated
Mechanism synergy 10 Combined effects accurate

Apache Score: 52/60 (86.7%)

Key Validated Behaviors

  1. Effort differentiation (4.12× median): High-loyalty contributors give ~4× more effort
  2. Phase-appropriate loyalty: Loyalty ceilings match historical observations
  3. Team size effects: Larger teams show expected per-capita reduction
  4. Free-riding baseline: Matches theoretical predictions (99.7% accuracy)

Example: Cross-Phase Comparison

import coopetition_gym
import numpy as np

phases = ["emergence", "growth", "maturity", "evolution"]
results = {}

for phase in phases: env = coopetition_gym.make("ApacheProject-v0", phase=phase)
    obs, info = env.reset(seed=42)

    # Cooperative strategy
    for step in range(60): actions = np.full(env.n_agents, 30.0)
        obs, rewards, terminated, truncated, info = env.step(actions)
        if terminated or truncated: break

    results[phase] = {
        "team_size": env.n_agents,
        "max_loyalty": info['phase_loyalty'],
        "mean_loyalty": info['mean_loyalty'],
        "team_output": info['team_output'],
    }
    env.close()

for phase, data in results.items(): print(f"{phase}: {data}")

References

  1. Pant, V. & Yu, E. (2026). Computational Foundations for Strategic Coopetition: Formalizing Collective Action and Loyalty. arXiv:2601.16237
  2. Mockus, A., Fielding, R., & Herbsleb, J. (2002). Two Case Studies of Open Source Software Development: Apache and Mozilla. ACM TOSEM.
  3. Lerner, J. & Tirole, J. (2002). Some Simple Economics of Open Source. Journal of Industrial Economics.