Skip to content

Latest commit

 

History

History
30 lines (23 loc) · 1.62 KB

File metadata and controls

30 lines (23 loc) · 1.62 KB

findings.md — Research & Discoveries

Project: Mastery Reinforcement & Retry Decision Engine

Updated: 2026-03-04


Constraints Confirmed

  • 100% offline — no external API calls
  • Python 3 standard library only (no pip installs required)
  • Docker for containerization and execution
  • CLI output only (no GUI/web)

Key Design Discoveries

Weighted Average Logic

  • Why 70/30? The 70% weight on the most recent score is intentional: it rewards genuine improvement while being skeptical of historical failure. A learner who scored 30, 35 then suddenly 80 should NOT pass after 3 attempts (Hard Stop overrides).
  • Edge Case: With only 1 score, use it directly — no weighted calc possible.

The False Positive Problem

  • The "Lucky Guesser" archetype is the critical risk. A learner scoring [30, 35, 80] has a weighted avg of 80*0.7 + 35*0.3 = 56 + 10.5 = 66.5. This WOULD normally push to RETRY zone, but since attempt_number=3, Hard Stop triggers PIVOT. This is safe.
  • If the sequence were [60, 70, 80] at attempt 3, weighted avg = 80*0.7 + 70*0.3 = 56 + 21 = 77. This is RETRY zone → Hard Stop → PIVOT. The skeptic rule holds.
  • Mastery (>= 80 weighted avg) at attempt 3 is still PROGRESS — the gate wasn't compromised, the learner genuinely earned it.

Docker Strategy

  • Single-stage build using python:3.11-slim base image — minimal footprint
  • Entry point runs mastery_engine.py which simulates all archetypes
  • No volumes or bind mounts needed — all data is generated programmatically

Pending Research

  • Validate Python 3.11-slim availability on user machine (will check during Link phase)