{
  "what_if": "No cheat: the leaders' public record, read entry by entry",
  "blocked": null,
  "python_version": "3.12.3",
  "base_mean_edge": 109.27254289017199,
  "base_unseen": {
    "mean_edge": 98.28063388702982,
    "retail_edge": 126.01907227628422,
    "arb_edge": -27.73843838925235
  },
  "diff": {
    "mean": 0.0,
    "standard_error": 0.0,
    "simulations": 4000,
    "seed_start": 50000,
    "retail": 0.0,
    "arb": 0.0
  },
  "by_spread": [
    {
      "spread": 1,
      "runs": 1016,
      "diff": 0.0,
      "standard_error": 0.0
    },
    {
      "spread": 2,
      "runs": 998,
      "diff": 0.0,
      "standard_error": 0.0
    },
    {
      "spread": 3,
      "runs": 1017,
      "diff": 0.0,
      "standard_error": 0.0
    },
    {
      "spread": 4,
      "runs": 969,
      "diff": 0.0,
      "standard_error": 0.0
    }
  ],
  "how": "No simulator or strategy rule changed for the no-cheat what-if. A separate legal allocation experiment is reported only in attempt. Cloned https://github.com/danrobinson/prediction-market-challenge at b244530ee2a6247b5f930830eeca9dd950da0e7e into workspace .cache, extracted base code from https://api.imdnotdead.fun/artifacts/ecbc9087b4f238fc1b5b0d52f55af073f676c038bfb6756172b1e51480ac1741, ran python3 -m orderbook_pm_challenge run base.py --simulations 600 --seed-start 1000 --workers 4 --json, then 4000 simulations starting 50000. Made a separate unchanged simulator copy under .cache and ran the 4000-seed command independently again. Compared result objects seed by seed. The public chronology and distribution evidence are in tried; all chart metrics come from the submissions API. No source code for either leader is public in these responses.",
  "clean": "Clean no-op control: no extra information, capital, timing advantage, liquidity, fill priority or change to RNG, scoring, fees, collateral, retail or arbitrage. diff.mean=0 is mandated by this observational item and independently verified; it is not a causal ceiling for an implementable leader mechanism. Missing evidence: leader code, per-seed public outcomes and shared-bin data. Published histograms are marginal, differ in boundaries, and nine total only 199 despite 200 successes. Paired winning bins and precise architectural changes cannot be supplied honestly. Nothing was fabricated; missing observations were not filled.",
  "realizable": "The public record supports two outcome targets, not two known mechanisms. Against IMD, Littoral v16 saves 3.432385 of arbitrage loss while adding 0.606881 retail; KING adds 3.643624 retail with only 0.560570 lower arbitrage loss. These are descriptive gaps, not transferable or additive causal gains. Most informative within-author steps are Ponxter v15->v16 (3.264583 less arbitrage loss for 2.113718 less retail) and Ethan Predictor->KING (8.015714 less arbitrage loss with 0.576370 more retail). A legal strategy can use observable fills, competitor quote changes, cash/inventory and remaining time to condition which quotes and inventory commitments it makes, improving adverse-selection/retail tradeoffs. Costs include foregone benign retail fills, reduced queue priority after cancellation, collateral tied in inventory, and model errors. The base already uses probes, a sub-tick probability filter, sticky orders and low-price YES stock, so merely naming these is not a new mechanism. The simulator calls the strategy before advancing the latent process and before arbitrage; reacting to the same step jump is unavailable. Conditional collateral allocation/inventory shadow prices may be worth an ablation, but neither the public aggregates nor code lengths prove leaders use them. A separate real attempt tests a deliberately simple allocation mechanism: reserve part of the currently allowed cash for planned uncovered asks before bids and stock acquisition spend it. Selection uses only training seeds; this heuristic may sacrifice beneficial bid fills and cheap YES accumulation. Its patch and paired unseen results are in attempt. It is a hypothesis motivated by the channel tradeoff, not a known leader implementation. The actual ask-priority trial returned paired unseen mean -0.0022317855786250116 (standard error 0.0017917106041472873). This is one realized heuristic trial; the attainable ceiling remains unmeasured. The fixed real trial changed edges only for competitor spreads [1]; 3981 of 4000 run edges were identical. It gives no evidence of a material gain from this specific priority rule. Its small realized effect does not bound the gains from a full joint inventory/collateral policy.",
  "attempt": {
    "name": "Reserve cash for desired uncovered asks before bid spending",
    "selected_variant": "askreserve0.5",
    "selection_rule": "Larger mean on the same 600 training seeds among two nonzero ask-priority variants; evaluate this one even if worse than baseline.",
    "training_base_mean_edge": 100.61264031170599,
    "training": [
      {
        "variant": "askreserve0.5",
        "mean_edge": 100.607027586005,
        "diff": -0.00561272570100004,
        "standard_error": 0.004770607244918786,
        "simulations": 600,
        "seed_start": 4000,
        "runs_better": 0,
        "runs_worse": 3,
        "runs_equal": 597
      },
      {
        "variant": "askreserve1.0",
        "mean_edge": 100.59610511924834,
        "diff": -0.01653519245766665,
        "standard_error": 0.013880774230214112,
        "simulations": 600,
        "seed_start": 4000,
        "runs_better": 1,
        "runs_worse": 5,
        "runs_equal": 594
      }
    ],
    "mean_edge": 98.27840210145119,
    "retail_edge": 126.01768430505916,
    "arb_edge": -27.73928220360577,
    "mean": -0.0022317855786250116,
    "standard_error": 0.0017917106041472873,
    "simulations": 4000,
    "seed_start": 50000,
    "retail": -0.0013879712250750105,
    "arb": -0.000843814353425006,
    "by_spread": [
      {
        "spread": 1,
        "runs": 1016,
        "diff": -0.008786557396161463,
        "standard_error": 0.007052548824121391
      },
      {
        "spread": 2,
        "runs": 998,
        "diff": 0.0,
        "standard_error": 0.0
      },
      {
        "spread": 3,
        "runs": 1017,
        "diff": 0.0,
        "standard_error": 0.0
      },
      {
        "spread": 4,
        "runs": 969,
        "diff": 0.0,
        "standard_error": 0.0
      }
    ],
    "failures": 0,
    "runs_better": 5,
    "runs_worse": 14,
    "runs_equal": 3981,
    "patch": "--- base.py\n+++ askreserve0.5.py\n@@ -256,7 +256,22 @@\n         add_b = (qb - kb) if (kb < qb * P['keep_lo'] and bid_t != cb) else 0.0\n         add_a = (qa - ka) if (ka < qa * P['keep_lo'] and ask_t != ca) else 0.0\n \n-        budget = P['fb'] * max(free, 0.0)\n+        # Legal cash allocation experiment: reserve funding for desired naked asks\n+        # before bid and inventory-acquisition orders consume their shared cash.\n+        # Uses only this on_step state's cash, orders, inventory and remaining time.\n+        ask_cash = 0.0\n+        if ask_t <= 99 and add_a >= 0.01:\n+            ask_x = ask_t / 100.0\n+            covered_kept = sum(o.remaining_quantity - min(self.res.get(o.order_id, 0.0), o.remaining_quantity)\n+                               for o in asks if o.order_id in kept_ids)\n+            available_yes = max(0.0, Y - covered_kept - 0.1)\n+            desired_unc = max(0.0, add_a - available_yes)\n+            allowed_unc_cash = max(0.0, free - (1 - P['fu']) * max(state.cash, 0.0)) if ask_x >= P['xu'] else 0.0\n+            if H < P['Hu'] and phat < 0.05:\n+                allowed_unc_cash = max(0.0, free * (1 - H / P['Hu']))\n+            allowed_unc_cash = min(allowed_unc_cash, max(0.0, free - C.get('reserve', 0.0) * min(1.0, H / C.get('release', 50.0))))\n+            ask_cash = 0.5 * min(desired_unc * (1 - ask_x), allowed_unc_cash)\n+        budget = P['fb'] * max(free - ask_cash, 0.0)\n         cost = add_b * bid_t / 100.0 + qam * amt / 100.0\n         if cost > budget:\n             f = budget / cost\n",
    "clean": "Real strategy code only; no simulator changes or privileged fields. Reserve amount uses existing state cash, own orders, YES inventory, planned ask, collateral constraints and steps remaining. Keeps existing ask risk caps and reserve rule, and reduces the cash budget available to bids and inventory acquisition. It is a priority heuristic, not a solved joint dynamic programme. Training uses only 4000..4599 within the authorized 4000..7999 range; evaluated the preselected candidate once on 50000..53999. No other candidate evaluated on unseen seeds."
  },
  "tried": [
    "SOURCE AND METHOD: chronological order is createdAt ascending within each author, not version number or leaderboard rank. Public numbers are displayed API precision, not raw unrounded run statistics. Public meanArbEdge is negative; arb_loss=-meanArbEdge; positive delta_arb_loss is worse. codeLength is the API field, not verified UTF-8 byte length. All proposed mechanism labels are low-confidence inferences, not source-code findings.",
    "PUBLIC METRICS SOURCE: https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions; 19 Ponxter entries, 2 Ethanann0424 entries, and IMD seat 1968. Every selected submission reports successCount=200 and failureCount=0; code=null. The entry table includes all 21 requested leader records.",
    "CHART COLUMNS: author | UTC timestamp | strategy | id | mean_edge | retail_edge | arb_loss | codeLength | delta_mean | delta_retail | delta_arb_loss | delta_codeLength | bin_win_result | interpretation. Deltas compare the immediately previous chronological record from the SAME author. First-entry deltas are unavailable, not zero.",
    "@Ponxter | 2026-10-25T20:30:07.821154Z | Littoral v1 | ff498a42-b2c9-4446-9fb6-baddf2095741 | 78.062935 | 108.587124 | 30.524189 | 26293 | NA | NA | NA | NA | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | First observed full strategy, no earlier public baseline; mechanism unknown.",
    "@Ponxter | 2026-10-25T20:58:27.109116Z | Littoral v2 | 282ad8f0-230b-4a21-a8d0-761d0c5d7ef7 | 84.730245 | 118.834901 | 34.104656 | 27304 | 6.667310 | 10.247777 | 3.580467 | 1011 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely more aggressive retail access or size: retail improves strongly while arbitrage loss rises; 1011 extra code-length units do not identify implementation.",
    "@Ponxter | 2026-10-26T00:40:21.237807Z | Littoral v3 | 93a981d1-7d76-41eb-8456-4fcd1dc9686e | 88.944961 | 121.158636 | 32.213675 | 34031 | 4.214716 | 2.323735 | -1.890981 | 6727 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely improved conditional quoting or sizing: both channels improve and code grows; fair-value, inventory and risk logic cannot be distinguished.",
    "@Ponxter | 2026-10-26T12:06:02.775276Z | Littoral v4 | 8dce73e5-394a-489e-968c-2777a33a9950 | 100.560941 | 123.231657 | 22.670716 | 51018 | 11.615980 | 2.073021 | -9.542959 | 16987 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Large improvement chiefly in adverse selection, consistent with better stale-quote protection, probability inference or risk sizing. Large code growth suggests added logic, not proof of any particular model.",
    "@Ponxter | 2026-10-31T22:03:25.631270Z | Predictor | 68b451dc-966f-4981-ba24-6763725be753 | 104.65111 | 125.892476 | 21.241366 | 56312 | 4.090169 | 2.660819 | -1.429350 | 5294 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely mixed retail-access and adverse-selection refinement; both channels improve.",
    "@Ponxter | 2026-11-01T01:18:45.698164Z | Predictor | d9137644-6e65-470e-90bc-067ac0ee261e | 105.64624 | 127.673466 | 22.027227 | 56548 | 0.99513 | 1.780990 | 0.785861 | 236 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely increased quoting aggressiveness or retail size: extra retail edge exceeds extra arbitrage loss. Only 236 code-length units added.",
    "@Ponxter | 2026-11-01T13:00:38.750911Z | Littoral v7 | 441981bb-71ad-4f14-8d8a-559e0672a02d | 105.538464 | 126.597928 | 21.059464 | 59309 | -0.107776 | -1.075538 | -0.967763 | 2761 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely defensive risk/inventory adjustment: sacrifices slightly more retail than it saves on arbitrage; no net progress.",
    "@Ponxter | 2026-11-02T12:42:12.255578Z | Littoral v9 | c7b638ee-3139-493c-9c09-ef21dcf1cbc6 | 105.530453 | 126.371787 | 20.841334 | 60516 | -0.008011 | -0.226141 | -0.218130 | 1207 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely risk/retail rebalance with negligible net effect; added code is not evidence of a superior architecture.",
    "@Ponxter | 2026-11-02T13:43:12.455482Z | Littoral v8 | a432f47e-921d-40d8-a403-e434b6099b99 | 105.768229 | 126.520266 | 20.752036 | 59455 | 0.237776 | 0.148479 | -0.089298 | -1061 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Earlier version label submitted later; smaller implementation and tiny improvements in both channels. Cannot separate variant selection from evaluation variability.",
    "@Ponxter | 2026-11-02T14:00:38.607463Z | Littoral v9 | ba5b42cc-ee9a-42fe-a4fe-5cc34853da30 | 105.461559 | 126.257152 | 20.795594 | 60516 | -0.306670 | -0.263114 | 0.043558 | 1061 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Revisited v9 after v8; code length matches the earlier v9, but code identity is unavailable. Likely replay or small variant; do not treat the changed score as a proven new mechanism.",
    "@Ponxter | 2026-11-02T15:18:28.809606Z | Captain 400 | 54235375-4fef-4e8a-9153-6aa04511ed23 | -38.899972 | 1.849817 | 40.749788 | 1211 | -144.361531 | -124.407335 | 19.954194 | -59305 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely diagnostic or distinct minimal strategy, not an incremental Littoral improvement: most code and nearly all retail edge disappear.",
    "@Ponxter | 2026-11-02T15:22:37.288059Z | Captain 400s | 040d05e1-041d-44fd-a906-0acc9450fe84 | -38.899972 | 1.849817 | 40.749788 | 1329 | 0.000000 | 0.000000 | 0.000000 | 118 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely behavior-preserving wrapper or inactive change to the diagnostic: 118 extra code-length units, identical displayed scores. Equal scores do not establish identical source.",
    "@Ponxter | 2026-11-02T15:27:53.699456Z | Littoral v10 | 91445558-a2e8-479c-a249-777f1ffba111 | 105.79686 | 129.793647 | 23.996787 | 59459 | 144.696832 | 127.943830 | -16.753001 | 58130 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Returns from diagnostic to the large Littoral branch; chronological jump is not a useful innovation estimate. Versus preceding Littoral v9, likely more aggressive retail participation at greater arbitrage cost.",
    "@Ponxter | 2026-11-02T21:48:54.950982Z | Littoral v11 | d6a6b61c-b2f4-4598-ae16-7b97abd2f6fc | 106.124815 | 129.412167 | 23.287352 | 59745 | 0.327955 | -0.381480 | -0.709435 | 286 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely defensive size/quote adjustment: smaller retail edge, greater reduction in arbitrage loss.",
    "@Ponxter | 2026-11-02T23:06:03.530514Z | Littoral v12 | 4537360c-6474-4286-9d79-94a8461deb01 | 108.325355 | 129.740703 | 21.415348 | 59118 | 2.200540 | 0.328536 | -1.872004 | -627 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely targeted adverse-selection improvement or simplification: fewer code-length units, gains in both channels, mostly arbitrage protection.",
    "@Ponxter | 2026-11-03T00:28:45.980680Z | Littoral v13 | 1bbd32c7-cc92-4ef8-8d6f-99e7a7b30972 | 106.722381 | 127.886013 | 21.163632 | 60225 | -1.602974 | -1.854690 | -0.251716 | 1107 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely overly defensive participation/inventory change: minor arbitrage saving fails to offset much larger retail loss.",
    "@Ponxter | 2026-11-03T01:22:27.085912Z | Littoral v14 | 43e0c6c7-a364-4f98-a9ce-21bc24b473b8 | 108.560502 | 129.862974 | 21.302472 | 59150 | 1.838121 | 1.976961 | 0.138840 | -1075 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely reversal or relaxation of v13 defensive changes: restores retail with little additional arbitrage loss; code length nearly returns to v12.",
    "@Ponxter | 2026-11-03T02:39:15.631035Z | Littoral v15 | 0a69e43a-83cc-4600-87eb-737901542c23 | 108.276351 | 129.848609 | 21.572258 | 59150 | -0.284151 | -0.014365 | 0.269786 | 0 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Equal code length and small score changes are consistent with constants, equal-length edits or evaluation variability. No public evidence of a structural innovation.",
    "@Ponxter | 2026-11-03T12:38:22.752578Z | Littoral v16 | 247570ef-6d9f-46c3-be85-c1187ad38563 | 109.427216 | 127.734891 | 18.307675 | 59197 | 1.150865 | -2.113718 | -3.264583 | 47 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely more selective quoting or inventory/risk gating: substantial arbitrage saving despite retail sacrifice and only 47 extra code-length units. Could be a threshold change; public aggregates cannot prove jump prediction or a new estimator.",
    "@Ethanann0424 | 2026-10-07T08:03:46.192467Z | Predictor | e08206c1-aa25-4bce-9a77-e9dbb61c9e3f | 101.00006 | 130.195264 | 29.195204 | 37326 | NA | NA | NA | NA | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | First observed Ethan entry; strong retail extraction with high adverse selection. No earlier public baseline.",
    "@Ethanann0424 | 2026-10-07T10:50:40.598572Z | KING | 5093d4c8-53d0-44e3-919a-cdf53bbdd25c | 109.592145 | 130.771634 | 21.17949 | 134648 | 8.592085 | 0.576370 | -8.015714 | 97322 | No identifiable paired winning bins (unpaired marginal histograms; bin boundaries vary). | Likely added conditional risk/inference logic or precomputation: most gain is reduced adverse selection with retail preserved and major code growth. This is low-confidence classification; tables, comments and formatting can also grow files.",
    "BIN COMPARISON LIMIT: A distribution bin is an edge interval, not a competitor spread, seed or regime. Its value field supplies no run identities and must not be treated as an observed within-bin mean. The API does not provide shared bin boundaries across submissions. Comparing the kth bin or rebucketing its entire count at its midpoint fabricates information. Neither matching a seed across entries nor a per-bin edge difference against IMD can be recovered. Therefore winning bins are unavailable for ALL 21 entries; no histogram is claimed to show paired wins.",
    "HISTOGRAM CHART: the following exact published [label,value,count] triples preserve the API data, without uniform-within-bin interpolation. 17 representative detail endpoints fit the 20-source budget including rules, base and submission list. Five omitted histograms add no identifiable paired-win information; all five entries and their aggregate channel/length changes are in the complete chronological chart above.",
    "HIST15 @Ethanann0424 Predictor https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/e08206c1-aa25-4bce-9a77-e9dbb61c9e3f count_sum=199 reported_successes=200 [[\"-81 to 10\",-35.32,61],[\"10 to 101\",55.87,87],[\"101 to 193\",147.07,18],[\"193 to 284\",238.26,11],[\"284 to 375\",329.46,9],[\"375 to 466\",420.65,3],[\"466 to 557\",511.85,6],[\"557 to 649\",603.04,3],[\"649 to 740\",694.24,0],[\"740 to 831\",785.43,0],[\"831 to 922\",876.63,0],[\"922 to 1013\",967.82,1],[\"1013 to 1105\",1059.02,0],[\"1105 to 1196\",1150.21,0],[\"1196 to 1287\",1241.41,0]]",
    "HIST15 @Ethanann0424 KING https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/5093d4c8-53d0-44e3-919a-cdf53bbdd25c count_sum=199 reported_successes=200 [[\"-19 to 75\",28.12,133],[\"75 to 169\",122.31,26],[\"169 to 264\",216.49,15],[\"264 to 358\",310.68,9],[\"358 to 452\",404.87,5],[\"452 to 546\",499.05,4],[\"546 to 640\",593.24,6],[\"640 to 735\",687.43,0],[\"735 to 829\",781.62,0],[\"829 to 923\",875.8,0],[\"923 to 1017\",969.99,0],[\"1017 to 1111\",1064.18,1],[\"1111 to 1205\",1158.36,0],[\"1205 to 1300\",1252.55,0],[\"1300 to 1394\",1346.74,0]]",
    "HIST15 @Ponxter Littoral v1 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/ff498a42-b2c9-4446-9fb6-baddf2095741 count_sum=199 reported_successes=200 [[\"-38 to 46\",4.42,128],[\"46 to 131\",88.48,35],[\"131 to 215\",172.54,16],[\"215 to 299\",256.61,7],[\"299 to 383\",340.67,5],[\"383 to 467\",424.73,4],[\"467 to 551\",508.79,3],[\"551 to 635\",592.85,0],[\"635 to 719\",676.92,0],[\"719 to 803\",760.98,0],[\"803 to 887\",845.04,1],[\"887 to 971\",929.1,0],[\"971 to 1055\",1013.16,0],[\"1055 to 1139\",1097.22,0],[\"1139 to 1223\",1181.29,0]]",
    "HIST15 @Ponxter Littoral v2 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/282ad8f0-230b-4a21-a8d0-761d0c5d7ef7 count_sum=200 reported_successes=200 [[\"-306 to -204\",-255.21,1],[\"-204 to -102\",-153.24,0],[\"-102 to -0\",-51.27,42],[\"-0 to 102\",50.69,110],[\"102 to 204\",152.66,21],[\"204 to 306\",254.63,8],[\"306 to 408\",356.59,6],[\"408 to 510\",458.56,8],[\"510 to 612\",560.53,2],[\"612 to 713\",662.5,0],[\"713 to 815\",764.46,0],[\"815 to 917\",866.43,1],[\"917 to 1019\",968.4,0],[\"1019 to 1121\",1070.37,0],[\"1121 to 1223\",1172.33,1]]",
    "HIST15 @Ponxter Littoral v3 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/93a981d1-7d76-41eb-8456-4fcd1dc9686e count_sum=199 reported_successes=200 [[\"-550 to -428\",-489.22,1],[\"-428 to -306\",-366.98,0],[\"-306 to -184\",-244.74,0],[\"-184 to -61\",-122.51,0],[\"-61 to 61\",-0.27,132],[\"61 to 183\",121.97,36],[\"183 to 305\",244.2,12],[\"305 to 428\",366.44,8],[\"428 to 550\",488.68,7],[\"550 to 672\",610.92,2],[\"672 to 794\",733.15,0],[\"794 to 917\",855.39,1],[\"917 to 1039\",977.63,0],[\"1039 to 1161\",1099.86,0],[\"1161 to 1283\",1222.1,0]]",
    "HIST15 @Ponxter Littoral v4 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/8dce73e5-394a-489e-968c-2777a33a9950 count_sum=200 reported_successes=200 [[\"-373 to -256\",-314.48,1],[\"-256 to -139\",-197.25,0],[\"-139 to -21\",-80.03,0],[\"-21 to 96\",37.19,143],[\"96 to 213\",154.41,24],[\"213 to 330\",271.63,12],[\"330 to 447\",388.86,8],[\"447 to 565\",506.08,8],[\"565 to 682\",623.3,2],[\"682 to 799\",740.52,0],[\"799 to 916\",857.74,1],[\"916 to 1034\",974.97,0],[\"1034 to 1151\",1092.19,0],[\"1151 to 1268\",1209.41,0],[\"1268 to 1385\",1326.63,1]]",
    "HIST15 @Ponxter Littoral v9 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/c7b638ee-3139-493c-9c09-ef21dcf1cbc6 count_sum=199 reported_successes=200 [[\"-404 to -281\",-342.57,1],[\"-281 to -157\",-218.93,0],[\"-157 to -33\",-95.28,0],[\"-33 to 90\",28.37,141],[\"90 to 214\",152.02,25],[\"214 to 337\",275.66,13],[\"337 to 461\",399.31,10],[\"461 to 585\",522.96,5],[\"585 to 708\",646.61,3],[\"708 to 832\",770.25,0],[\"832 to 956\",893.9,0],[\"956 to 1079\",1017.55,1],[\"1079 to 1203\",1141.2,0],[\"1203 to 1327\",1264.84,0],[\"1327 to 1450\",1388.49,0]]",
    "HIST15 @Ponxter Littoral v8 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/a432f47e-921d-40d8-a403-e434b6099b99 count_sum=199 reported_successes=200 [[\"-406 to -282\",-344.11,1],[\"-282 to -159\",-220.76,0],[\"-159 to -36\",-97.41,0],[\"-36 to 88\",25.95,140],[\"88 to 211\",149.3,26],[\"211 to 334\",272.66,13],[\"334 to 458\",396.01,10],[\"458 to 581\",519.37,6],[\"581 to 704\",642.72,2],[\"704 to 828\",766.08,0],[\"828 to 951\",889.43,1],[\"951 to 1074\",1012.78,0],[\"1074 to 1198\",1136.14,0],[\"1198 to 1321\",1259.49,0],[\"1321 to 1445\",1382.85,0]]",
    "HIST15 @Ponxter Captain 400 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/54235375-4fef-4e8a-9153-6aa04511ed23 count_sum=199 reported_successes=200 [[\"-104 to -97\",-100.34,2],[\"-97 to -89\",-93.09,15],[\"-89 to -82\",-85.84,16],[\"-82 to -75\",-78.6,13],[\"-75 to -68\",-71.35,11],[\"-68 to -60\",-64.1,9],[\"-60 to -53\",-56.85,7],[\"-53 to -46\",-49.61,9],[\"-46 to -39\",-42.36,12],[\"-39 to -31\",-35.11,10],[\"-31 to -24\",-27.86,15],[\"-24 to -17\",-20.62,4],[\"-17 to -10\",-13.37,3],[\"-10 to -2\",-6.12,36],[\"-2 to 5\",1.13,37]]",
    "HIST15 @Ponxter Captain 400s https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/040d05e1-041d-44fd-a906-0acc9450fe84 count_sum=199 reported_successes=200 [[\"-104 to -97\",-100.34,2],[\"-97 to -89\",-93.09,15],[\"-89 to -82\",-85.84,16],[\"-82 to -75\",-78.6,13],[\"-75 to -68\",-71.35,11],[\"-68 to -60\",-64.1,9],[\"-60 to -53\",-56.85,7],[\"-53 to -46\",-49.61,9],[\"-46 to -39\",-42.36,12],[\"-39 to -31\",-35.11,10],[\"-31 to -24\",-27.86,15],[\"-24 to -17\",-20.62,4],[\"-17 to -10\",-13.37,3],[\"-10 to -2\",-6.12,36],[\"-2 to 5\",1.13,37]]",
    "HIST15 @Ponxter Littoral v10 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/91445558-a2e8-479c-a249-777f1ffba111 count_sum=200 reported_successes=200 [[\"-441 to -314\",-377.27,1],[\"-314 to -187\",-250.7,0],[\"-187 to -61\",-124.13,0],[\"-61 to 66\",2.44,130],[\"66 to 192\",129.01,35],[\"192 to 319\",255.58,14],[\"319 to 445\",382.15,9],[\"445 to 572\",508.72,5],[\"572 to 699\",635.29,4],[\"699 to 825\",761.86,0],[\"825 to 952\",888.44,0],[\"952 to 1078\",1015.01,1],[\"1078 to 1205\",1141.58,0],[\"1205 to 1331\",1268.15,0],[\"1331 to 1458\",1394.72,1]]",
    "HIST15 @Ponxter Littoral v11 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/d6a6b61c-b2f4-4598-ae16-7b97abd2f6fc count_sum=200 reported_successes=200 [[\"-441 to -314\",-377.27,1],[\"-314 to -187\",-250.7,0],[\"-187 to -61\",-124.13,0],[\"-61 to 66\",2.44,131],[\"66 to 192\",129.01,34],[\"192 to 319\",255.58,14],[\"319 to 445\",382.15,9],[\"445 to 572\",508.72,5],[\"572 to 699\",635.29,4],[\"699 to 825\",761.86,0],[\"825 to 952\",888.44,0],[\"952 to 1078\",1015.01,1],[\"1078 to 1205\",1141.58,0],[\"1205 to 1331\",1268.15,0],[\"1331 to 1458\",1394.72,1]]",
    "HIST15 @Ponxter Littoral v12 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/4537360c-6474-4286-9d79-94a8461deb01 count_sum=200 reported_successes=200 [[\"-67 to 35\",-15.92,99],[\"35 to 137\",85.73,53],[\"137 to 238\",187.38,16],[\"238 to 340\",289.03,13],[\"340 to 442\",390.68,6],[\"442 to 543\",492.33,6],[\"543 to 645\",593.98,5],[\"645 to 746\",695.63,0],[\"746 to 848\",797.28,0],[\"848 to 950\",898.93,0],[\"950 to 1051\",1000.58,1],[\"1051 to 1153\",1102.23,0],[\"1153 to 1255\",1203.88,0],[\"1255 to 1356\",1305.53,0],[\"1356 to 1458\",1407.18,1]]",
    "HIST15 @Ponxter Littoral v13 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/1bbd32c7-cc92-4ef8-8d6f-99e7a7b30972 count_sum=199 reported_successes=200 [[\"-79 to 23\",-27.93,77],[\"23 to 125\",73.88,74],[\"125 to 227\",175.69,18],[\"227 to 328\",277.5,11],[\"328 to 430\",379.31,8],[\"430 to 532\",481.12,4],[\"532 to 634\",582.93,6],[\"634 to 736\",684.73,0],[\"736 to 837\",786.54,0],[\"837 to 939\",888.35,0],[\"939 to 1041\",990.16,1],[\"1041 to 1143\",1091.97,0],[\"1143 to 1245\",1193.78,0],[\"1245 to 1346\",1295.58,0],[\"1346 to 1448\",1397.39,0]]",
    "HIST15 @Ponxter Littoral v15 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/0a69e43a-83cc-4600-87eb-737901542c23 count_sum=200 reported_successes=200 [[\"-54 to 46\",-4.0,109],[\"46 to 147\",96.89,47],[\"147 to 248\",197.78,12],[\"248 to 349\",298.67,13],[\"349 to 450\",399.56,9],[\"450 to 551\",500.46,4],[\"551 to 652\",601.35,4],[\"652 to 753\",702.24,0],[\"753 to 854\",803.13,0],[\"854 to 954\",904.02,0],[\"954 to 1055\",1004.92,1],[\"1055 to 1156\",1105.81,0],[\"1156 to 1257\",1206.7,0],[\"1257 to 1358\",1307.59,0],[\"1358 to 1459\",1408.48,1]]",
    "HIST15 @Ponxter Littoral v16 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/247570ef-6d9f-46c3-be85-c1187ad38563 count_sum=200 reported_successes=200 [[\"-59 to 41\",-9.27,104],[\"41 to 141\",90.92,47],[\"141 to 241\",191.11,17],[\"241 to 341\",291.3,13],[\"341 to 442\",391.5,6],[\"442 to 542\",491.69,6],[\"542 to 642\",591.88,4],[\"642 to 742\",692.07,1],[\"742 to 842\",792.26,0],[\"842 to 943\",892.45,0],[\"943 to 1043\",992.65,1],[\"1043 to 1143\",1092.84,0],[\"1143 to 1243\",1193.03,0],[\"1243 to 1343\",1293.22,0],[\"1343 to 1444\",1393.41,1]]",
    "HIST15 @builtbyimd IMD seat 1968 https://www.paradigm.xyz/puzzles/api/v1/prediction-market/submissions/05c81652-ae1e-426f-9fdd-1f2e361cc255 count_sum=200 reported_successes=200 [[\"-101 to -10\",-55.36,3],[\"-10 to 81\",35.28,131],[\"81 to 171\",125.92,30],[\"171 to 262\",216.56,10],[\"262 to 353\",307.21,10],[\"353 to 443\",397.85,4],[\"443 to 534\",488.49,5],[\"534 to 624\",579.13,4],[\"624 to 715\",669.77,1],[\"715 to 806\",760.41,0],[\"806 to 896\",851.05,0],[\"896 to 987\",941.69,1],[\"987 to 1078\",1032.33,0],[\"1078 to 1168\",1122.97,0],[\"1168 to 1259\",1213.61,1]]",
    "CONSERVATIVE RANGE TEST: IMD published bins envelope=[-102.0,1260.0] after padding each integer label endpoint by 1.0 to allow for integer rounding. A competitor bin entirely above the envelope beats every IMD observation represented in its published histogram, irrespective of seed matching; a bin entirely below loses to every such observation. This is a comparison of the observed marginal samples, not a matched-regime or population claim. No within-bin interpolation is used.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v16 247570ef-6d9f-46c3-be85-c1187ad38563: above all represented IMD runs=['1343 to 1444 (1 runs)']; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v15 0a69e43a-83cc-4600-87eb-737901542c23: above all represented IMD runs=['1358 to 1459 (1 runs)']; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v13 1bbd32c7-cc92-4ef8-8d6f-99e7a7b30972: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v12 4537360c-6474-4286-9d79-94a8461deb01: above all represented IMD runs=['1356 to 1458 (1 runs)']; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v11 d6a6b61c-b2f4-4598-ae16-7b97abd2f6fc: above all represented IMD runs=['1331 to 1458 (1 runs)']; below all represented IMD runs=['-441 to -314 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v10 91445558-a2e8-479c-a249-777f1ffba111: above all represented IMD runs=['1331 to 1458 (1 runs)']; below all represented IMD runs=['-441 to -314 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Captain 400s 040d05e1-041d-44fd-a906-0acc9450fe84: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Captain 400 54235375-4fef-4e8a-9153-6aa04511ed23: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v8 a432f47e-921d-40d8-a403-e434b6099b99: above all represented IMD runs=[]; below all represented IMD runs=['-406 to -282 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v9 c7b638ee-3139-493c-9c09-ef21dcf1cbc6: above all represented IMD runs=[]; below all represented IMD runs=['-404 to -281 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v4 8dce73e5-394a-489e-968c-2777a33a9950: above all represented IMD runs=['1268 to 1385 (1 runs)']; below all represented IMD runs=['-373 to -256 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v3 93a981d1-7d76-41eb-8456-4fcd1dc9686e: above all represented IMD runs=[]; below all represented IMD runs=['-550 to -428 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v2 282ad8f0-230b-4a21-a8d0-761d0c5d7ef7: above all represented IMD runs=[]; below all represented IMD runs=['-306 to -204 (1 runs)']. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ponxter Littoral v1 ff498a42-b2c9-4446-9fb6-baddf2095741: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ethanann0424 KING 5093d4c8-53d0-44e3-919a-cdf53bbdd25c: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "CERTAIN OUTSIDE-RANGE BINS @Ethanann0424 Predictor e08206c1-aa25-4bce-9a77-e9dbb61c9e3f: above all represented IMD runs=[]; below all represented IMD runs=[]. All remaining bins: paired win status unknown. Missing observations are unclassified.",
    "HISTOGRAM DATA GAP: nine inspected endpoints have count_sum=199 versus successCount=200: 1bbd32c7-cc92-4ef8-8d6f-99e7a7b30972, 040d05e1-041d-44fd-a906-0acc9450fe84, 54235375-4fef-4e8a-9153-6aa04511ed23, a432f47e-921d-40d8-a403-e434b6099b99, c7b638ee-3139-493c-9c09-ef21dcf1cbc6, 93a981d1-7d76-41eb-8456-4fcd1dc9686e, ff498a42-b2c9-4446-9fb6-baddf2095741, 5093d4c8-53d0-44e3-919a-cdf53bbdd25c, e08206c1-aa25-4bce-9a77-e9dbb61c9e3f. The cause is unverified. Do not add a synthetic observation or assume that it is the maximum. The other reproduced histograms total 200.",
    "LEADER VERSUS IMD: @Ponxter Littoral v16 delta_mean=4.039266 delta_retail=0.606881 reduction_in_arb_loss=3.432385. Difference is between published aggregate records, not a paired causal improvement. Components can differ from total by 0.000001 because API values are rounded.",
    "LEADER VERSUS IMD: @Ethanann0424 KING delta_mean=4.204195 delta_retail=3.643624 reduction_in_arb_loss=0.56057. Difference is between published aggregate records, not a paired causal improvement. Components can differ from total by 0.000001 because API values are rounded.",
    "PUBLIC PATH TOTAL @Ponxter: first-to-last delta_mean=31.364281 delta_retail=19.147767 reduction_in_arb_loss=12.216514 delta_codeLength=32904. Descriptive endpoints, not a causal decomposition across matched seeds.",
    "PUBLIC PATH TOTAL @Ethanann0424: first-to-last delta_mean=8.592085 delta_retail=0.576370 reduction_in_arb_loss=8.015714 delta_codeLength=97322. Descriptive endpoints, not a causal decomposition across matched seeds.",
    "BRANCH CONTROL v10 versus most recent non-diagnostic v9: delta_mean=0.335301 delta_retail=3.536495 delta_arb_loss=3.201193. The chronological +144.696832 rebound from Captain 400s is not an innovation gain.",
    "REPRODUCTION: Python 3.12.3, simulator commit b244530ee2a6247b5f930830eeca9dd950da0e7e, strategy SHA256 baef955ec10728839f7c3357016ec9eb0550e7ec689bbd5171b1cdd5c9e37273. Extracted only the artifact code field; no constants changed; 4 workers accepted, sandbox disabled, no dependencies installed.",
    "BASE CHECK seeds 1000..1599: {\"mean_edge\": 109.27254289017199, \"retail_edge\": 135.356137274445, \"arb_edge\": -26.083594384290333}; 600 successes, 0 failures; reference=109.27254289017199; measured-minus-reference=0.0",
    "BASE UNSEEN seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; reference=98.28063388702982; measured-minus-reference=0.0",
    "NO-CHEAT INDEPENDENT REPLAY seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; all 4000 full SimulationResult objects exactly equal; 18 simulator/strategy/project source files byte-identical. Paired diff uses sample stdev(B-A)/sqrt(n), with sample variance denominator n-1. This zero is a control, NOT a measured ceiling on improvements inferred from leaders.",
    "REPRODUCTION: Python 3.12.3, simulator commit b244530ee2a6247b5f930830eeca9dd950da0e7e, strategy SHA256 baef955ec10728839f7c3357016ec9eb0550e7ec689bbd5171b1cdd5c9e37273. Extracted only the artifact code field; no constants changed; 4 workers accepted, sandbox disabled, no dependencies installed.",
    "BASE CHECK seeds 1000..1599: {\"mean_edge\": 109.27254289017199, \"retail_edge\": 135.356137274445, \"arb_edge\": -26.083594384290333}; 600 successes, 0 failures; reference=109.27254289017199; measured-minus-reference=0.0",
    "BASE UNSEEN seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; reference=98.28063388702982; measured-minus-reference=0.0",
    "NO-CHEAT INDEPENDENT REPLAY seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; all 4000 full SimulationResult objects exactly equal; 18 simulator/strategy/project source files byte-identical. Paired diff uses sample stdev(B-A)/sqrt(n), with sample variance denominator n-1. This zero is a control, NOT a measured ceiling on improvements inferred from leaders.",
    "REPRODUCTION: Python 3.12.3, simulator commit b244530ee2a6247b5f930830eeca9dd950da0e7e, strategy SHA256 baef955ec10728839f7c3357016ec9eb0550e7ec689bbd5171b1cdd5c9e37273. Extracted only the artifact code field; no constants changed; 4 workers accepted, sandbox disabled, no dependencies installed.",
    "BASE CHECK seeds 1000..1599: {\"mean_edge\": 109.27254289017199, \"retail_edge\": 135.356137274445, \"arb_edge\": -26.083594384290333}; 600 successes, 0 failures; reference=109.27254289017199; measured-minus-reference=0.0",
    "BASE UNSEEN seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; reference=98.28063388702982; measured-minus-reference=0.0",
    "NO-CHEAT INDEPENDENT REPLAY seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; all 4000 full SimulationResult objects exactly equal; 18 simulator/strategy/project source files byte-identical. Paired diff uses sample stdev(B-A)/sqrt(n), with sample variance denominator n-1. This zero is a control, NOT a measured ceiling on improvements inferred from leaders.",
    "REPRODUCTION: Python 3.12.3, simulator commit b244530ee2a6247b5f930830eeca9dd950da0e7e, strategy SHA256 baef955ec10728839f7c3357016ec9eb0550e7ec689bbd5171b1cdd5c9e37273. Extracted only the artifact code field; no constants changed; 4 workers accepted, sandbox disabled, no dependencies installed.",
    "BASE CHECK seeds 1000..1599: {\"mean_edge\": 109.27254289017199, \"retail_edge\": 135.356137274445, \"arb_edge\": -26.083594384290333}; 600 successes, 0 failures; reference=109.27254289017199; measured-minus-reference=0.0",
    "BASE UNSEEN seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; reference=98.28063388702982; measured-minus-reference=0.0",
    "NO-CHEAT INDEPENDENT REPLAY seeds 50000..53999: {\"mean_edge\": 98.28063388702982, \"retail_edge\": 126.01907227628422, \"arb_edge\": -27.73843838925235}; 4000 successes, 0 failures; all 4000 full SimulationResult objects exactly equal; 18 simulator/strategy/project source files byte-identical. Paired diff uses sample stdev(B-A)/sqrt(n), with sample variance denominator n-1. This zero is a control, NOT a measured ceiling on improvements inferred from leaders.",
    "REAL ATTEMPT TRAINING: [{\"variant\": \"askreserve0.5\", \"mean_edge\": 100.607027586005, \"diff\": -0.00561272570100004, \"standard_error\": 0.004770607244918786, \"simulations\": 600, \"seed_start\": 4000, \"runs_better\": 0, \"runs_worse\": 3, \"runs_equal\": 597}, {\"variant\": \"askreserve1.0\", \"mean_edge\": 100.59610511924834, \"diff\": -0.01653519245766665, \"standard_error\": 0.013880774230214112, \"simulations\": 600, \"seed_start\": 4000, \"runs_better\": 1, \"runs_worse\": 5, \"runs_equal\": 594}]; baseline=100.61264031170599; selection=askreserve0.5. Both nonzero candidates and baseline use identical 600 seeds starting at 4000; zero failures. Selection rule fixed before unseen evaluation and evaluates the better nonzero variant even if both trail baseline.",
    "REAL ATTEMPT UNSEEN: {\"name\": \"Reserve cash for desired uncovered asks before bid spending\", \"selected_variant\": \"askreserve0.5\", \"selection_rule\": \"Larger mean on the same 600 training seeds among two nonzero ask-priority variants; evaluate this one even if worse than baseline.\", \"training_base_mean_edge\": 100.61264031170599, \"mean_edge\": 98.27840210145119, \"retail_edge\": 126.01768430505916, \"arb_edge\": -27.73928220360577, \"mean\": -0.0022317855786250116, \"standard_error\": 0.0017917106041472873, \"simulations\": 4000, \"seed_start\": 50000, \"retail\": -0.0013879712250750105, \"arb\": -0.000843814353425006, \"by_spread\": [{\"spread\": 1, \"runs\": 1016, \"diff\": -0.008786557396161463, \"standard_error\": 0.007052548824121391}, {\"spread\": 2, \"runs\": 998, \"diff\": 0.0, \"standard_error\": 0.0}, {\"spread\": 3, \"runs\": 1017, \"diff\": 0.0, \"standard_error\": 0.0}, {\"spread\": 4, \"runs\": 969, \"diff\": 0.0, \"standard_error\": 0.0}], \"failures\": 0, \"runs_better\": 5, \"runs_worse\": 14, \"runs_equal\": 3981}. The no-cheat diff above is unaffected; this separate attempt is not claimed to reproduce leader code or establish a ceiling.",
    "REAL ATTEMPT INTERPRETATION: The fixed real trial changed edges only for competitor spreads [1]; 3981 of 4000 run edges were identical. It gives no evidence of a material gain from this specific priority rule. Its small realized effect does not bound the gains from a full joint inventory/collateral policy."
  ],
  "next": "The ask-priority variant changed 19 of 4000 unseen run edges and had paired mean -0.0022317855786250116 with SE 0.0017917106041472873. This tests only the simple cash-reservation priority, not the full value of joint collateral optimization. First build a seed-paired attribution harness on seeds 4000..7999 that logs retail misses, arbitrage losses, inventory and collateral at each quote decision, split by spread, time remaining and low-probability inventory scarcity. Compare a dynamic marginal value of collateral across both sides against the existing independent size rules; freeze it before evaluating seeds 50000..53999. Treat this as a hypothesis, not a reconstruction of either leader. Seek public paired seed results or leader source before claiming any winning histogram bins or jump-prediction mechanism. The zero no-op result closes no strategy direction."
}