EXP-PENT70-R9

series member

Pentagon 70rds — N=5, 14 rds/agent, rotation=9

2026-03-10 L3 level paper3 1 runs $0.01

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

Testing why experiments with 7 agents are more stable than with 5. Hypothesis: it is not about the number of agents, but the number of rounds per agent. We give 5 agents the same number of rounds per agent (14) as the seven had.

Predictions we made before running

0/2 confirmed

▶ Technical details

Research hypothesis

VARIANCE REDUCTION TEST. N=5 with 70 rounds (14 rds/agent) instead of 50 (10 rds/agent). Tests whether N=7's lower variance (SD=0.080, CV=16.2%) vs N=5's higher variance (SD=0.137, CV=33.7%) is driven by rounds/agent (14 vs 10) rather than N itself.

If rounds/agent drives variance: SD should drop from 0.137 to <0.10. If N drives variance: SD should stay >0.12 regardless of more rounds.

Experimental setup

Type: simple

Condition	Parameters
PENT70_LIVE	persona: PERSONA, interaction: LIVE, n_agents: 5, n_rounds: 70, note: N=5, 70 rounds (14/agent). Rotation=9.

Factors: n_agents (5) × rounds_per_agent (14)

Parameters

n_agents: 5
n_rounds: 70
n_runs_per_condition: 1
model: gemini-2.5-flash-lite
temperature: 0.9
scheduler: round_robin

Trophic Ratios by Condition

Mean trophic ratio per agent across runs. Error bars = ±1 std dev. Higher TR = more upstream (exporter).