{"key": "values7-1", "track": "B", "status": "candidate_not_better_than_base_on_partial_tuning; confirmation and holdout NOT run (time)", "mechanism": "Independent copy 1 of the posterior-value idea. Belief update/arb call/rival and phase rules are hash-pinned ThrLattice thr=0.9 (thrpol.py sha256 663965707d4f20f81ab0ade8ed510742a330605fd9e889d426db1d920abc8958) via the earlier valuepolicy.py. Only the quote lookup changes (postval.py): shortlist of 25 (ask,bid) pairs around the base optimum (steps of step*{-2..2}, base first, ties keep base, no parabola); <=5 posterior price nodes (arb point mass + <=4 equal-mass buckets of the retail cut; 5 buckets of the lattice otherwise); score = sum node mass*[multilinear passive VN(own disp, rival disp, a-e, b+e) + kappa(1-h)IG], VN regenerated by pinned dp.py (110 sweeps, resumable DP_STATE, 4 bounded commands, gain 536.8206, last diff 1.88e-7, float64). Gaussian ablation (mode=gauss) uses the identical kernel, actions and branches but replaces the retail cut by a moment-matched Gauss-Hermite rule.", "gate_replacement": "The impossible equality (argmax of interpolated values == interpolated optimal fees off-grid; previously 6.909 bps) is dropped. Replaced by: grid/non-tie argmax and pi0/1 branch-state checks (prior artifact check_endpoints/check_branches, inherited, not rerun here), and off-grid independent evaluation of the SAME actions (pure-Python vs numba kernel, max abs U diff in audit below) plus quantified fee drift vs base lookup.", "frozen_settings": {"status": "no candidate cleared tuning; NOT frozen for promotion", "tried": [{"mode": "post", "step": 0.5, "kn": 5}, {"mode": "gauss", "step": 0.5, "kn": 5}], "not_run": ["post step=1.0", "4th setting"], "qm": "qm_k1e-5.json", "thr": 0.9}, "tuning": {"requested_range": [64000, 64255], "actual_range": [64000, 64063], "n_actual": 64, "reason": "run.py chunking and a lost ~8 min command (my mistake: a timed-out compound command) left too little time; no seeds imputed", "post_vs_base": {"n": 64, "mean_diff": -0.10741059474303594, "se": 0.06026858971538138, "sum": -6.8742780635543, "sumsq": 15.383815613360257, "worst5": [[64029, -1.107259368193752], [64051, -1.0693334874464426], [64039, -1.0356650802210652], [64006, -0.9444385779216873], [64000, -0.7922840344639326]]}, "gauss_ablation_vs_base": {"n": 64, "mean_diff": -0.03276954652202768, "se": 0.05071860519568771, "sum": -2.0972509774097716, "sumsq": 10.440549676672633, "worst5": [[64054, -0.867948147304844], [64059, -0.7281078219286883], [64051, -0.6547269565302258], [64005, -0.6488079256088213], [64008, -0.5741724813733526]]}, "arm_stats": {"post": {"n": 64, "mean_edge": 524.8837270838203, "first_trades": 450330, "first_trade_misread": 0.28732040947749427, "price_rms_bps": 13.177927164152646}, "gauss": {"n": 64, "mean_edge": 524.9583681320413, "first_trades": 449909, "first_trade_misread": 0.28724697661082577, "price_rms_bps": 13.204849567509907}, "base": {"n": 64, "mean_edge": 524.9911376785634, "first_trades": 449582, "first_trade_misread": 0.29089020467901294, "price_rms_bps": 13.342293540835108}}, "posterior_minus_gauss_on_these_seeds": -0.07464104822100825, "reading": "Both within ~2 se of zero and negative; posterior gain over matched Gaussian-kernel cannot be separated from noise at n=64; interpolation/shortlist-kernel gain (gauss vs base) -0.033 se 0.051."}, "confirmation": {"status": "NOT RUN", "requested_range": [64256, 64511], "actual_n": 0}, "holdout": {"status": "NOT RUN", "requested_range": [10000, 12047], "actual_n": 0, "note": "no holdout number claimed; no incremental gain claimed; no promotion"}, "incremental_gain_over_base": {"measured": null, "tuning_n64_post": -0.10741059474303594, "inherited_threshold_gain": "not measured here; base already includes thr=0.9", "promote": false}, "checks": {"one_seed_64000_check_post": {"seeds": [64000, 64000], "off_vs_base_max_fee_diff_bps": 0.0, "replay_max_fee_diff_bps": 0.0, "callbacks_off": 9777, "audit_all": {"n": 6624, "same_action_rate": 0.9461050724637681, "max_abs_dU_chosen": 0.000805979358223701, "mean_regret_red_under_full": 6.4529722218744e-07, "max_regret": 6.970615402224559e-05, "max_fee_diff_red_vs_full_bps": 4.998576255951237, "mean_nfull": 26.79453502415459, "mean_nred": 4.931914251207729, "gain_over_base_action_mean": 3.3877196949170095e-05, "gain_full_mean": 3.212340451053073e-05}, "audit_ambiguous_first": {"n": 5602, "same_action_rate": 0.9501963584434131, "max_abs_dU_chosen": 0.000805979358223701, "mean_regret_red_under_full": 4.936694934286504e-07, "max_regret": 6.970615402224559e-05, "max_fee_diff_red_vs_full_bps": 4.998576255951237, "mean_nfull": 26.899678686183506, "mean_nred": 4.982863263120314, "gain_over_base_action_mean": 3.7350283434381976e-05, "gain_full_mean": 3.5546514801612236e-05}, "audit_first": {"n": 6623, "same_action_rate": 0.9460969349237506, "max_abs_dU_chosen": 0.000805979358223701, "mean_regret_red_under_full": 6.453946549553982e-07, "max_regret": 6.970615402224559e-05, "max_fee_diff_red_vs_full_bps": 4.998576255951237, "mean_nfull": 26.79842971463083, "mean_nred": 4.9325079269213346, "gain_over_base_action_mean": 3.388231203250834e-05, "gain_full_mean": 3.212825479054138e-05}, "py_vs_jit_max_abs_U": 0.0, "audit_calls": 6624, "fee_drift_vs_base_bps": {"n": 9717, "mean_abs": 0.7278613494238896, "p50_abs": 0.0, "p99_abs": 4.974633443201036, "max_abs": 4.99875020830709, "frac_changed": 0.49284758670371515}}, "disabled_matches_base": "mode=off vs ThrLattice: max fee diff 0.0 bps over 9777 callbacks (seed 64000)", "causal_replay": "fresh object fed only the six afterSwap inputs reproduces recorded fees, max diff 0.0 bps (seed 64000); policy never receives flag/price/rival/run params", "approximation_error": "reduced (~4.9 nodes) vs full (~26.8 hyps) posterior on 6624 audited calls: same action in 94.6%; max chosen-U difference 8.1e-4; mean regret 6.5e-7; max fee difference 5.0 bps (one shortlist half-step=2.5 bps units x ...).", "fee_drift_vs_base": "see check.fee_drift_vs_base_bps (candidate fees vs base lookup)", "numba_parity": "pure-Python and numba kernels agree on all audited calls (max abs U diff 0.0); fallback pure Python exists"}, "limitations": ["Only 64 tuning seeds, no confirmation/holdout: no conclusion about 0.3 bar beyond \"no sign of gain\".", "Prior-step rival uncertainty collapsed to scalar (inherited approximation).", "Offsets chosen against moment-based kappa IG as in the base.", "Elapsed time mostly spent regenerating VN (~45 min on 8 CPUs) and one lost command.", "Misread rate shown (~29%) uses run.py definition (is_arb pi>0.9 per thr) and equals base within noise; it is not the 16% of thr=0.5 lattice."], "cost": {"extra_storage_slots": 0, "per_call_ops": "<=5 nodes x 25 actions x 16 multilinear corners = 2000 value reads + IG lookups; ~6000 gas of reads at 3 gas lower bound ignoring arithmetic. UNMEASURED estimate, no EVM run; VN tables themselves (~0.5 GB float64) are far over any contract limit, so a port is not viable as is."}, "code_encoding": "base64(zlib(USTAR tar)); sha256 over uncompressed tar", "code_files": ["postval.py", "check_post.py"], "code_sha256": "99e6e866eff38380e30ebef1cbaeb9cb566420a7a691ec4c0967505fb929a1a5", "code_zb64": "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", "reproduction": "Needs prior artifact 2640642f... (valuepolicy.py, actionvalues.py, regenerate.py) + pinned solved folder; regenerate VN with DP_SAVEVN=1 DP_STATE=.. DP_SWEEPS=32 (4 commands); python run.py postval:PostPolicy 64000 64 out.json qm=qm_k1e-5.json values=<prefix> mode=post step=0.5; pair.py vs thrpol:ThrLattice thr=0.9.", "seed_ledger": {"engine_seeds": 0, "replica_tuning": [64000, 64063], "confirmation": null, "holdout": null}}