{"key": "fresh3-1", "idea": "Cost-weighted arb/retail call: keep the lattice posterior as is, but read the first own trade of a step as the arb (phase 0, no rival-spot update) only when its posterior arb probability exceeds 0.9 instead of 0.5, because mistaking a retail fill for the arb costs more than the reverse.", "settings_tried": [{"setting": "thr=0.3 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": -0.3442, "se": 0.0259}, {"setting": "thr=0.4 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": -0.1783, "se": 0.0214}, {"setting": "thr=0.6 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.1239, "se": 0.0185}, {"setting": "thr=0.7 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.217, "se": 0.0306}, {"setting": "thr=0.8 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.2908, "se": 0.0342}, {"setting": "thr=0.85 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.309, "se": 0.0353}, {"setting": "thr=0.88 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.3154, "se": 0.0357}, {"setting": "thr=0.9 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.3362, "se": 0.0369}, {"setting": "thr=0.92 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.3328, "se": 0.0381}, {"setting": "thr=0.95 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.2987, "se": 0.0362}, {"setting": "thr=0.98 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.2575, "se": 0.0391}, {"setting": "thr=0.995 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.1979, "se": 0.046}, {"setting": "thr=1.01 vs thr=0.5 (latt.py), seeds 60000..60255", "n": 256, "mean_diff": 0.1661, "se": 0.0506}], "confirm_paired": {"n": 256, "mean_diff": 0.2941, "se": 0.0297, "setting": "thr=0.9 vs thr=0.5, seeds 60256..60511"}, "holdout_paired": {"n": 2048, "mean_diff": 0.3229, "se": 0.0107, "sum": 661.282652, "sumsq": 692.740899, "worst5": [[11487, -1.912], [10942, -1.608], [10982, -1.446], [11380, -1.237], [10990, -1.15]], "reference": "ref/lattice_qm.json", "seeds": "10000..12047, four 512-seed commands, thr=0.9", "vs_HA": "+4.586 se 0.038 (lattice_qm +4.263)"}, "misread_rate": {"base": 0.1622, "mine": 0.2888, "note": "mine counts the thresholded call; the belief itself is identical to the base"}, "price_error_bps": {"base": 13.64, "mine": 13.28}, "gas_estimate": "about 0 extra: one comparison against a constant replaces pi>0.5 in the phase branch", "code_zb64": "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", "code_sha256": "f6504da3b5013be52bd556059e8744e8799f715e2334dbc9e772e057ce8c9287", "notes": "Why: the call only picks phase-0 vs phase-1/2 tables and whether a fill updates the rival spot; the posterior already holds both hypotheses. Wrongly calling a retail fill the arb loses the rival sync and quotes phase-0 (wide unwind) offsets, which costs more than the reverse, so the best cut sits well above 0.5. Tuning was flat from 0.85 to 0.95 (+0.30 to +0.34) and fell off on both sides; a cut at 1.0 (never call the arb) still gains +0.17. The misread count goes up (16.2% to 28.9%) while edge goes up, so misread rate is the wrong target. Runs: numba runner fast_thr.py (fast_replica.py of the fresh3 artifact with a thr option, 0 diff to latt.py at thr=0.5 on 32 seeds), and thrpol.py through run.py Blind matched it to 1e-9 on seeds 10000..10005 at thr=0.9. The multiprocessing pool could not start on this seat; runs were single-process. Untested: stacking it with the pi-weighted log-blend (+0.18), which acts on the same ambiguous trades, so the gains may overlap. Next: put the cut on the blend weight, or make the cut depend on fill size and sd. Holdout +0.32 se 0.011 clears the 0.3 bar for Track A only just."}