{ "credit": "Taylor (gtmcknight), github.com/gtmcknight/amm-sim@4812a6382311624e1506aca131ca35c378f930d9 research/solved: replica, Gaussian filter, fee rule, qm tables, arb/retail tell", "mech": "lookahead: additive forecast variance used only for the fee-table lookup", "already_present": { "value": false, "where": "Read leanbase.py, stalerival.py, real.py and tellpol.py from reference_zb64 of https://api.imd.fun/artifacts/efb8e653f8fd061cecf3e92b186470c9c96c5188523364b60ed101e9672869aa", "explanation": "real.RealPolicy.on_trade adds SIG2 for elapsed timestamp steps, then real.RealPolicy.lookup receives sqrt(current posterior variance). stalerival.Policy.on_trade repeats that current-variance lookup after its rival pull. leanbase overrides the tell; tellpol changes the arb likelihood. real.lookup has optional multiplicative sdscale (default 1), not additive forward-time variance. No tested nearest variant was needed. This is not a claim that the offline optimized tables ignore future trades." }, "definition": { "formula": "sigma0=0.000945; sd_lookup_bps=10000*sqrt(v+h*sigma0^2), h>=0. Keep lookup coordinates w,w2,phase,synced unchanged. Keep stored m,v unchanged by the forecast itself; changed fees naturally affect later observations.", "parameters": { "h": "Fixed proxy for expected waiting time to the next event, measured in simulation timestamp steps; globally chosen, not inferred from hidden per-run sigma or arrival rate.", "sigma0": 0.000945 }, "state_kept": "No additional dynamic state. h is a configuration scalar in F[7]; it can be a compile-time constant in a contract. Uses existing posterior v and fixed SIG2 only. No future prices, rival reserves, order counts or hidden flags.", "scope": "One compiled afterSwap lookup changed. Initial fees, tables, filter, tell, rival pull, simulator and hook remain as in the supplied replica; hook disabled.", "selection": { "seed_start": 64000, "seed_end": 65023, "round_size": 128, "settings_count": 12, "n_steps": 10000, "baseline": { "a0": 0.35, "ainf": 0.35, "tau": 2.0 }, "stop_rule": "At the end of each cumulative 128-seed round, drop immediately if mean_diff+3*se<0.", "selection_rule": "Highest final paired mean among survivors; if all dropped, highest stopped mean; ties smaller h.", "se": "Sample standard deviation of paired per-seed edge differences (ddof=1) divided by sqrt(n).", "rounds": { "0.001": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.006539551441939739, "se": 0.0075999833112352765, "seeds_positive": 59, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.006872507892280888, "se": 0.005482150266581194, "seeds_positive": 120, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": -0.0004913880882668012, "se": 0.004968244457830251, "seeds_positive": 191, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.0023441800328785067, "se": 0.0043148567132109865, "seeds_positive": 261, "dropped": false }, { "seed_end": 64639, "n": 640, "mean_diff": -0.004412261597782052, "se": 0.00387892666461422, "seeds_positive": 320, "dropped": false }, { "seed_end": 64767, "n": 768, "mean_diff": -0.0019429736358185974, "se": 0.003513895485756568, "seeds_positive": 384, "dropped": false }, { "seed_end": 64895, "n": 896, "mean_diff": -0.0004437612766556219, "se": 0.003341580642203549, "seeds_positive": 455, "dropped": false }, { "seed_end": 65023, "n": 1024, "mean_diff": -0.00022780953631007517, "se": 0.0031345830901480516, "seeds_positive": 521, "dropped": false } ], "0.003": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.01481919466155901, "se": 0.011966936068324321, "seeds_positive": 59, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.01222143729394265, "se": 0.007511488011668341, "seeds_positive": 115, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": -0.006701725388935357, "se": 0.006124173957116787, "seeds_positive": 177, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.00695479151370193, "se": 0.005198889420853475, "seeds_positive": 231, "dropped": false }, { "seed_end": 64639, "n": 640, "mean_diff": -0.008301802596777819, "se": 0.004610942069680705, "seeds_positive": 297, "dropped": false }, { "seed_end": 64767, "n": 768, "mean_diff": -0.008000732968190963, "se": 0.004172272603635148, "seeds_positive": 359, "dropped": false }, { "seed_end": 64895, "n": 896, "mean_diff": -0.007293458464227623, "se": 0.003973446642789197, "seeds_positive": 430, "dropped": false }, { "seed_end": 65023, "n": 1024, "mean_diff": -0.007285024539197105, "se": 0.00367735123141019, "seeds_positive": 493, "dropped": false } ], "0.01": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.01147729662105057, "se": 0.012729136408461754, "seeds_positive": 55, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.003516757918663682, "se": 0.008787545619025264, "seeds_positive": 116, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": 7.871558211336449e-05, "se": 0.006968376068980739, "seeds_positive": 182, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.00551675392310258, "se": 0.006045080807004302, "seeds_positive": 238, "dropped": false }, { "seed_end": 64639, "n": 640, "mean_diff": -0.009481205206355892, "se": 0.005515940948730486, "seeds_positive": 298, "dropped": false }, { "seed_end": 64767, "n": 768, "mean_diff": -0.008702062521647994, "se": 0.005283883577156164, "seeds_positive": 353, "dropped": false }, { "seed_end": 64895, "n": 896, "mean_diff": -0.0071696984483393044, "se": 0.0048522857461453205, "seeds_positive": 422, "dropped": false }, { "seed_end": 65023, "n": 1024, "mean_diff": -0.008647656019028671, "se": 0.004567532721199253, "seeds_positive": 475, "dropped": false } ], "0.02": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.008889263313765383, "se": 0.016069860957545225, "seeds_positive": 60, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.00016472600343808708, "se": 0.010483583165904743, "seeds_positive": 121, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": -0.0026690566367541337, "se": 0.008416222458670356, "seeds_positive": 185, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.008260733389533925, "se": 0.007188721377774975, "seeds_positive": 243, "dropped": false }, { "seed_end": 64639, "n": 640, "mean_diff": -0.011572198084446938, "se": 0.006421233908625367, "seeds_positive": 298, "dropped": false }, { "seed_end": 64767, "n": 768, "mean_diff": -0.010530300863257791, "se": 0.006159527723720133, "seeds_positive": 354, "dropped": false }, { "seed_end": 64895, "n": 896, "mean_diff": -0.009325459665023152, "se": 0.005619080845290178, "seeds_positive": 412, "dropped": false }, { "seed_end": 65023, "n": 1024, "mean_diff": -0.011023669008297043, "se": 0.00523397605086843, "seeds_positive": 459, "dropped": false } ], "0.04": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.025264556391231263, "se": 0.015694705069791304, "seeds_positive": 58, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.009806790705984003, "se": 0.01139281520232592, "seeds_positive": 126, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": -0.012650320699070386, "se": 0.009252082543380736, "seeds_positive": 185, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.016458448528052116, "se": 0.007950938028675355, "seeds_positive": 248, "dropped": false }, { "seed_end": 64639, "n": 640, "mean_diff": -0.01729715408721679, "se": 0.007053945269336478, "seeds_positive": 306, "dropped": false }, { "seed_end": 64767, "n": 768, "mean_diff": -0.018660388214923218, "se": 0.006634203974532996, "seeds_positive": 357, "dropped": false }, { "seed_end": 64895, "n": 896, "mean_diff": -0.01519619598539996, "se": 0.006130005565218898, "seeds_positive": 422, "dropped": false }, { "seed_end": 65023, "n": 1024, "mean_diff": -0.01566586893708588, "se": 0.00586408208729923, "seeds_positive": 475, "dropped": false } ], "0.08": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.03256714838205621, "se": 0.019469851668295627, "seeds_positive": 54, "dropped": false }, { "seed_end": 64255, "n": 256, "mean_diff": -0.017967736958262037, "se": 0.01362444570184652, "seeds_positive": 116, "dropped": false }, { "seed_end": 64383, "n": 384, "mean_diff": -0.02858732605794545, "se": 0.01105926938041847, "seeds_positive": 164, "dropped": false }, { "seed_end": 64511, "n": 512, "mean_diff": -0.031066397778732324, "se": 0.009506419860149648, "seeds_positive": 219, "dropped": true } ], "0.16": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.07047086187064, "se": 0.02305390868577803, "seeds_positive": 51, "dropped": true } ], "0.25": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.09418173406792052, "se": 0.027656476557755074, "seeds_positive": 52, "dropped": true } ], "0.4": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.1735434209459532, "se": 0.0345218191883362, "seeds_positive": 42, "dropped": true } ], "0.6": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.2934953511718845, "se": 0.04035647170854096, "seeds_positive": 32, "dropped": true } ], "1.0": [ { "seed_end": 64127, "n": 128, "mean_diff": -0.6510755355968616, "se": 0.05600499771223599, "seeds_positive": 17, "dropped": true } ], "2.0": [ { "seed_end": 64127, "n": 128, "mean_diff": -1.7325480894667957, "se": 0.0808882087220874, "seeds_positive": 6, "dropped": true } ] } }, "report_protocol": "Freeze chosen h using selection data only. Run each seed 10000..12047 once for each of four policies: LEAN, LEAN+h, AGE, AGE+h; 10000 steps each. AGE uses a0=0, ainf=0.7, tau=1. Same seeds paired within each comparison; no holdout retuning.", "sources": { "replica": "https://api.imd.fun/artifacts/a88ff982eba7e078bc6ccb1555b90a211c00c9fd76ec90a62ab13a4074f2c3d0", "reference": "https://api.imd.fun/artifacts/efb8e653f8fd061cecf3e92b186470c9c96c5188523364b60ed101e9672869aa", "tables": "https://raw.githubusercontent.com/gtmcknight/amm-sim/66604b3426d8e97a73e2c39940b9b4dece51cc73/research/solved/", "kit": "https://raw.githubusercontent.com/identity-md-launches/goal-3178-build-shared-evaluation-kit/45e7563d1a29ffca9d608fdf60c90bf2624d2d01/README.md" }, "source_sha256": "d9c758bb280fe24e09c039496f84ab0804173ff70cec24c332f8683f28efa582", "patched_source_sha256": "229cc9292f532220419cb5775259aac0220e35b4f07141875e95b51bec16438a", "patch_sha256": "cf4a3c81bb8caa21f674dca94c5c9a8c1e99961f16f8e31d828b79139a3413a9", "dependency_sha256": { "leanbase.py": "ba015769f7981d7798acbbc9aff8c0e0c59825bb1541d55bcdef49311d7a04bc", "stalerival.py": "d124a171d35b024daed840e5449d352b2b5e50a68ce98c8a9aaafaf3419d23da", "real.py": "6a4a0f0e7f8b43d38e5eaa9d8b2e847a0771aee7f89551b122fe89bfe7969b5f", "tellpol.py": "7d5961c45bec56292533e65607245ad119b8733c760a5bf28f76a2e07bc3960b", "lean_tell.py": "b70054f8c0561779e40c4c8df3f41ab08da05b9aef591816ac6c9034beb40a95", "tellref.py": "4867a33106ce2597863b8d7c8fb396288954b98f291714801fae8820f1f47aa3", "replica2.py": "54380c8755f4e42e9a0426759e918eacd2c5f915b7f04e5d0501c394c5846b40", "evalpol.py": "79a96efe8c543aee125c937554688b37ec2e7df44df4f3435265bba6eb012bf3", "par.py": "bf8a1a2e2a4bf9a23bf4622d726fd815702298ad1f9fa0dd17197230cce63ec1", "run.py": "91133e4ef485626226679d73349b5b4e374d9713ac5d7e9d36a39e1afd75933c", "pol_mid.json": "efbb5ee102db4d312f64565d9bf7232b6becbe02b624164f744a402bbbf2983b", "qm_k1e-5.json": "5ff53e42ca87390e5f65d1caa3f7a480e67a8212fb0670274dfd562b05ba1afd" }, "reproduction": "Decode patch_zb64 with base64 then zlib, apply unified diff to the original fastlean.py. With numpy==2.2.6, llvmlite==0.44.0, numba==0.61.2 on CPython 3.12: tabs=fastlean.load(amm_dir, a0=..., ainf=..., tau=..., lookahead=h); edges=fastlean.run_batch(tabs, range(seed0,seed0+n)), default n_steps=10000 and weights=None. Use the seed schedule, settings, stopping and selection rules above. Original tables and modules are pinned by dependency_sha256.", "validation": { "seeds": "63000..63007, disjoint from selection and report", "zero_h_bit_identical_to_unmodified_fastlean_both_baselines": true, "oracle": "Independent leanbase.Policy subclass overrides lookup to call parent lookup with 10000*sqrt(self.v+h*real.SIG2). Each callback is also replayed through the modified compiled on_trade; whole-run edges compared against run_batch.", "oracle_policy_seed_pairs": 16, "callbacks": 171362, "max_abs_edge_difference": 0.0, "max_abs_fee_difference_bps": 0.0, "arb_retail_decision_mismatches": 0, "invalid_parameters_rejected": true, "evidence": "Locally executed checks, not an independent certification; four-way tell labels were not separately re-certified." } }, "settings_tried": [ { "parameters": { "h": 0.001 }, "n": 1024, "mean_diff": -0.00022780953631007517, "se": 0.0031345830901480516, "seeds_positive": 521, "dropped_early": false }, { "parameters": { "h": 0.003 }, "n": 1024, "mean_diff": -0.007285024539197105, "se": 0.00367735123141019, "seeds_positive": 493, "dropped_early": false }, { "parameters": { "h": 0.01 }, "n": 1024, "mean_diff": -0.008647656019028671, "se": 0.004567532721199253, "seeds_positive": 475, "dropped_early": false }, { "parameters": { "h": 0.02 }, "n": 1024, "mean_diff": -0.011023669008297043, "se": 0.00523397605086843, "seeds_positive": 459, "dropped_early": false }, { "parameters": { "h": 0.04 }, "n": 1024, "mean_diff": -0.01566586893708588, "se": 0.00586408208729923, "seeds_positive": 475, "dropped_early": false }, { "parameters": { "h": 0.08 }, "n": 512, "mean_diff": -0.031066397778732324, "se": 0.009506419860149648, "seeds_positive": 219, "dropped_early": true }, { "parameters": { "h": 0.16 }, "n": 128, "mean_diff": -0.07047086187064, "se": 0.02305390868577803, "seeds_positive": 51, "dropped_early": true }, { "parameters": { "h": 0.25 }, "n": 128, "mean_diff": -0.09418173406792052, "se": 0.027656476557755074, "seeds_positive": 52, "dropped_early": true }, { "parameters": { "h": 0.4 }, "n": 128, "mean_diff": -0.1735434209459532, "se": 0.0345218191883362, "seeds_positive": 42, "dropped_early": true }, { "parameters": { "h": 0.6 }, "n": 128, "mean_diff": -0.2934953511718845, "se": 0.04035647170854096, "seeds_positive": 32, "dropped_early": true }, { "parameters": { "h": 1.0 }, "n": 128, "mean_diff": -0.6510755355968616, "se": 0.05600499771223599, "seeds_positive": 17, "dropped_early": true }, { "parameters": { "h": 2.0 }, "n": 128, "mean_diff": -1.7325480894667957, "se": 0.0808882087220874, "seeds_positive": 6, "dropped_early": true } ], "chosen": { "h": 0.001, "selection_mean_diff": -0.00022780953631007517, "reason": "Highest final paired selection mean among surviving settings; a nearly disabled forecast, not evidence for useful lookahead." }, "report": { "vs_lean": { "n": 2048, "mean_diff": 0.003492132563304662, "se": 0.002146116474430133, "seeds_positive": 1029, "seed_start": 10000, "seed_end": 12047, "baseline_mean_edge": 536.603390956178, "treatment_mean_edge": 536.6068830887414, "paired_difference_f64le_sha256": "5b71beeeaaaa5ef9f9f1231aebbc09101a5e8dc91b8e9b5caa959abb6eae7d2b" }, "on_age": { "n": 2048, "mean_diff": 0.0005059215832921293, "se": 0.0022894789176444662, "seeds_positive": 1044, "seed_start": 10000, "seed_end": 12047, "baseline_mean_edge": 536.6231002985178, "treatment_mean_edge": 536.623606220101, "paired_difference_f64le_sha256": "554f8656d8cf14d76ecb85983138f450e273903a6cf9d765140c0a67714b6ca1" } }, "promote": false, "contract_cost": { "storage_words": 0, "gas": { "measured_increment": null, "analysis": "With h compiled in, precompute h*SIG2; add that constant before the existing square root. No extra SLOAD/SSTORE, table or separate square root is intrinsically needed. Actual peak delta depends on fixed-point representation, sqrt implementation and changed table branches; no compiled Solidity gas claim.", "baseline_peak_from_brief": 208000, "limit": 245000, "headroom": 37000 }, "characters": { "measured_solidity_increment": null, "analysis": "A constant declaration and one addition inside the existing sqrt expression. Illustratively sqrt(v) -> sqrt(v + LOOKAHEAD_VAR) adds 16 characters at the expression site, excluding the constant declaration and unit conversion. This is not a completed Solidity port or a measured source size.", "baseline_from_brief": 47230, "limit": 50000, "headroom": 2770 }, "runtime_bytes": { "measured_increment": null, "baseline_from_brief": 22440, "limit": 24576, "headroom": 2136, "analysis": "Small arithmetic change with no new tables; actual compiler output unmeasured. Fit within any contract limit is not certified." } }, "patch_zb64": "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", "not_done": [ "No Solidity contract, compilation, bytecode/source-size measurement, gas or OOG checks; no build/check/engine pipeline because this bounded item is a replica experiment.", "No QLX5 or qt1 comparison, fresh-engine evaluation, or leaderboard submission. No claim of a +0.04 engine gain. No promotion pipeline unless both specified replica report means reach +0.03.", "No engine-faithful k recomputation fix: this intentionally uses the specified fast LEAN replica unchanged outside the lookup forecast. Engine transfer is untested.", "No full P/T/F/G/E certification; local zero-h and Python fee/decision/edge checks are described in definition.validation. Q-style selection/report seed ranges were followed.", "No other mechanisms or spent variants tested. No empirical arrival-rate estimate; h is a single global horizon proxy.", "Working environments and per-seed checkpoint files remain under test/scratch and are not delivered. Reproduction requires the pinned original artifact and table downloads; no executable build dependency is claimed to be included." ] }