id: dfa96c24f57d4c14bfdb546e402398ad
parent_id: 75332a8f491548b79d898be861c9fd5d
item_type: 1
item_id: f3c42f4435644e19b186265dd3fbf672
item_updated_time: 1787152545034
title_diff: "[]"
body_diff: "[{\"diffs\":[[1,\"# Improvement Ideas — Top 10 (2026-08-14)\\\nID: f3c42f4435644e19b186265dd3fbf672\\\nNotebook ID: 2c8da247905946c3aa19eb4936e16323\\\nCreated: 1786733906375\\\nUpdated: 1786789294456\\\nIs Todo: No\\\nTags: roadmap, gen5, rl, ideas\\\n\\\n---\\\n\\\n# Idea #8 PROPOSED: Opponent Mirrors + Decision-Time Rollouts (2026-08-19)\\\n\\\n**Status: IDEA — candidate for Gen 7 / post-v19. Proposed by Jan.**\\\n\\\n## Concept\\\nKeep per-opponent \\\"mirrors\\\" (behavioral models from observer stats) and, at decision\\\ntime, simulate how the hand might develop: sample opponent hole cards from the current\\\nrange posterior, sample their actions from an oracle behavioral model, roll the hand\\\nforward, and average EV per hero action.\\\n\\\n## Why the pieces now exist\\\n- **Mirror = observer stats → personality class**: already implemented for the\\\n  oracle-in-corrector attempt (class from stats, oracle row lookup). What's missing\\\n  is using it for FORWARD simulation instead of range correction.\\\n- **Range posterior**: RangeNet v19 P(hand|actions) per opponent (in progress).\\\n- **Behavioral oracle**: empirical P(action | strength-decile, facing-size, class)\\\n  from v19 equity_data (regen planned). Closes the loop:\\\n  posterior → sample hands → sample actions → rollout → EV(hero action).\\\n\\\n## Two known constraints (from 2026-08 findings)\\\n1. **Human transfer**: pool-bot P(call|strength) is flatter (2.7:1) than human (8:1) —\\\n   the oracle-in-corrector regression. For LIVE mirrors, blend pool prior with\\\n   live-observed frequencies per opponent (hand_history data) instead of raw oracle.\\\n2. **Cost**: decision-time rollouts multiply tree breadth × MC equity. Path: run\\\n   rollouts OFFLINE as label machinery, distill into an EV/action net (EquityNet\\\n   rollout-phase pattern, parked at Multiway-MAE 0.0448 — could be revived as the\\\n   distillation target with action-aware labels).\\\n\\\n## Suggested staging\\\n1. Oracle table regen from v19 (already planned)\\\n2. Rollout evaluator in sims vs pool (exact mirrors — no transfer issue on sim side)\\\n3. A/B as g56-style candidate; if wins → distill for live (net replaces runtime rollouts)\\\n\\\n---\\\n\\\n\"],[0,\"# Idea #\"]],\"start1\":0,\"start2\":0,\"length1\":8,\"length2\":2080}]"
metadata_diff: {"new":{},"deleted":[]}
encryption_cipher_text: 
encryption_applied: 0
updated_time: 2026-08-19T15:16:05.112Z
created_time: 2026-08-19T15:16:05.112Z
is_locked: 0
type_: 13