# Merge 07 of r7-t03-joint: continuing fee policy T3 by CMA-ES on the replica (items j061–j070)

All figures below are copied or computed from the ten item results under `.imd/reads/artifacts/item-j061 … item-j070`; every file's sha256 was recomputed and equals the one given in the assignment. Source for each row is the item named in it. Real-engine numbers are paired per seed, each strategy alone against the 30 bps VanillaStrategy, `amm_sim_rs.run_batch(X, VanillaStrategy, configs, 8)`.

## Headline

- Only 3 of 10 items ran a search (j061, j063, j070); the other 7 measured replica speed below 500 markets/s on their machine and, as their brief prescribed, ran **real-only** (T3 versus HA), no CMA-ES.
- The three searches gave replica holdout gains over T3 of +2.558, +2.564 and +2.536 (8,192 markets each), but over T3's own starting point on the same holdout the gain is only +0.025, +0.031 and +0.003 (over_base). The monitor curves are flat (about 2.58–2.63 over T for every checkpoint) after the very first monitor point, which was 2.5846 for all three at the start.
- On the real engine, the candidates beat T3 by only +0.013 to +0.064 per block (1,024 seeds each, se 0.0095–0.0138): roughly 0 to 4.6 se. They did not approach the +3.1 over HA the goal needs. T3 is ahead of HA by about +1.41 (pooled over the 7 real-only items, 7,168 seeds, se 0.012; see `real_only`), and the search candidates add only about +0.01 to +0.06 on top of that.
- No candidate had any OOG call; all afterSwap peaks are ≤ 245,000 (highest 238,247). All gates A, B and C that were run passed (Gate C not applicable/not run in real-only items).
- **Best valid search row (highest holdout over_T): j063** (+2.5641, se 0.0136), code sha256 `7e588fca…025b`. Full code and params at the bottom.

## Caveats on sources

The assignment flags sources that did not hold on re-read for j061 (2), j062 (1), j063 (1), j064 (1), j065 (2), j066 (1), j067 (3), j068 (1), j070 (1); j069 had none. The result files themselves contain no source list or URL citations apart from four `api.imd.fun/artifacts/<id>` input URLs in j066 (T3, HA, replica, CG artifacts); j067's notes mention a `sources.json` that is not in the result I was given. I therefore cannot tell which statements the flagged sources backed. What I did instead: every number here is a measurement reported inside the result (not a quote from a page), and the input hash checks they report (T3 `2333ecf2…`, code `cb6d9455…`, HA `2801efc8…`/`99e3a2ec…`, CG `ebbd90c6…`, replica `e8a7a2af…`) are consistent across all ten. I have not re-run any of the measurements. Treat the claims of j061, j062, j063, j064, j065, j066, j067, j068 and j070 as less certain than j069's, specifically wherever they rest on their provenance statements rather than on their own numbers. I did not find any numeric disagreement tied to a flagged source.

## `rows`

All numbers are strings. `null` = not run. Search rows: holdout seeds 40000000..40008191 (n=8192). `real` = candidate minus T3 for search rows, per block. For real-only rows the submitted code is T3 unchanged (sha256 cb6d9455…), and the real blocks are in `real_only`.

```json
[
 {
  "key": "j061",
  "mode": "search",
  "markets_per_s": "918.026",
  "generations": "289",
  "search_minutes": "51.3792",
  "holdout_over_T": {
   "mean": "2.55847",
   "se": "0.0137679"
  },
  "holdout_over_base": {
   "mean": "0.0253707",
   "se": "0.00391732"
  },
  "real_candidate_minus_T3": [
   {
    "seeds": "8940000..8941023",
    "n": "1024",
    "mean": "0.0423672",
    "se": "0.0107508",
    "sum": "43.384",
    "sumsq": "122.914"
   },
   {
    "seeds": "8942000..8943023",
    "n": "1024",
    "mean": "0.0337805",
    "se": "0.0111532",
    "sum": "34.5912",
    "sumsq": "131.477"
   }
  ],
  "afterSwap_peak": "237464",
  "OOG": "0",
  "valid": "true",
  "last_curve_point": {
   "minutes": "50.9911",
   "monitor_gain_over_T": "2.61252",
   "sigma": "0.206644"
  },
  "code_sha256": "07f7636112e1802376ee42bbb0fdf2fa96e85b20c1887c13f7a78ad2edf8f8b4",
  "setting_note": "setting ?: population 16, es_seed 861, sigma0 0.5"
 },
 {
  "key": "j062",
  "mode": "real-only",
  "markets_per_s": "42.2359",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "B"
 },
 {
  "key": "j063",
  "mode": "search",
  "markets_per_s": "635.132",
  "generations": "201",
  "search_minutes": "55.8528",
  "holdout_over_T": {
   "mean": "2.56412",
   "se": "0.0136057"
  },
  "holdout_over_base": {
   "mean": "0.0310249",
   "se": "0.00375996"
  },
  "real_candidate_minus_T3": [
   {
    "seeds": "8948000..8949023",
    "n": "1024",
    "mean": "0.0279043",
    "se": "0.0103535",
    "sum": "28.574",
    "sumsq": "113.089"
   },
   {
    "seeds": "8950000..8951023",
    "n": "1024",
    "mean": "0.0640293",
    "se": "0.0138272",
    "sum": "65.566",
    "sumsq": "204.482"
   }
  ],
  "afterSwap_peak": "238120",
  "OOG": "0",
  "valid": "true",
  "last_curve_point": {
   "minutes": "55.2283",
   "monitor_gain_over_T": "2.62786",
   "sigma": "0.192015"
  },
  "code_sha256": "7e588fca4d153358cdaa397eae44d9760d5192c222d83b36442aac037a95025b",
  "setting_note": "setting C: population 32, es_seed 863, sigma0 0.5"
 },
 {
  "key": "j064",
  "mode": "real-only",
  "markets_per_s": "124.561",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "see note"
 },
 {
  "key": "j065",
  "mode": "real-only",
  "markets_per_s": "104.387",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "D"
 },
 {
  "key": "j066",
  "mode": "real-only",
  "markets_per_s": "450.051",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "A"
 },
 {
  "key": "j067",
  "mode": "real-only",
  "markets_per_s": "120.684",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "E"
 },
 {
  "key": "j068",
  "mode": "real-only",
  "markets_per_s": "101.018",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "A"
 },
 {
  "key": "j069",
  "mode": "real-only",
  "markets_per_s": "180.48",
  "generations": null,
  "search_minutes": null,
  "holdout_over_T": null,
  "holdout_over_base": null,
  "real_candidate_minus_T3": null,
  "afterSwap_peak": null,
  "OOG": null,
  "valid": "true",
  "last_curve_point": null,
  "code_sha256": "cb6d945511e1e10af78baa64c2a6cb25806f0998403ce6440654bfc3c70092d9",
  "note": "real-only; no search, no candidate; code is T3 unchanged",
  "requested_setting": "see note"
 },
 {
  "key": "j070",
  "mode": "search",
  "markets_per_s": "677.284",
  "generations": "96",
  "search_minutes": "56.1628",
  "holdout_over_T": {
   "mean": "2.53587",
   "se": "0.0134736"
  },
  "holdout_over_base": {
   "mean": "0.00277191",
   "se": "0.0034307"
  },
  "real_candidate_minus_T3": [
   {
    "seeds": "8976000..8977023",
    "n": "1024",
    "mean": "0.01297",
    "se": "0.00954978",
    "sum": "13.2813",
    "sumsq": "95.7072"
   },
   {
    "seeds": "8978000..8979023",
    "n": "1024",
    "mean": "0.0130884",
    "se": "0.00987729",
    "sum": "13.4025",
    "sumsq": "102.376"
   }
  ],
  "afterSwap_peak": "238247",
  "OOG": "0",
  "valid": "true",
  "last_curve_point": {
   "minutes": "55.3378",
   "monitor_gain_over_T": "2.59243",
   "sigma": "0.191651"
  },
  "code_sha256": "4ced577f61ffb2b626c8c2304653d32f7244eaca75ebf7c88ce4fbe2d7a79a7f",
  "setting_note": "setting A: population 16, es_seed 870, sigma0 0.5"
 }
]
```

Notes on rows: j061 and j063 give each block's gas for the candidate; j063's second-block seeds and gas are in the source file. Peaks: j061 candidate 237,464 (T3 237,473), j063 candidate 238120 (T3 238108), j070 candidate 238,247 (T3 238,247). Retail/arb *edge* deltas are null in every real block (the engine wheel exposes only total edge, pnl and volumes); volume deltas below.

Candidate-minus-T3 volume (Y per market), from the real blocks:

- j061 seeds 8940000..8941023: retail volume 97.5922, arb volume 178.876; wins 559/1024
- j061 seeds 8942000..8943023: retail volume 97.1287, arb volume 180.12; wins 559/1024
- j063 seeds 8948000..8949023: retail volume 4.88592, arb volume 68.6653; wins 569/1024
- j063 seeds 8950000..8951023: retail volume 4.8915, arb volume 64.8838; wins 593/1024
- j070 seeds 8976000..8977023: retail volume 7.62527, arb volume 74.7339; wins 544/1024
- j070 seeds 8978000..8979023: retail volume 4.10162, arb volume 76.2357; wins 516/1024

Holdout (replica, n=8192): over_T / over_base: j061 +2.5585 se 0.0138 / +0.0254 se 0.0039 (retail +0.372, arb −0.347 over base); j063 +2.5641 se 0.0136 / +0.0310 se 0.0038; j070 +2.5359 se 0.0135 / +0.0028 se 0.0034 (within one se of zero).

Best monitored point: j061 2.6127 (se 0.0300) at 20.3 min, gen 116; j063 2.6327 at 45.1 min; j070 2.6013 at 26.6 min (before freeze rounding). Training budgets: j061 289 gens × 512 markets, pop 16; j063 201 gens × 256 markets, pop 32, 51,456 unique markets (setting C); j070 96 generations × 512, pop 16. j070's monitor curve ended below its start-to-5-min level (2.5924 at 55.3 min vs 2.5846 start), i.e. no clear continued improvement. The goal's premise that curves were still rising is not supported by these three curves: they are essentially flat from the first monitor.

Gates (search items): A (T3 port vs EVM) max diff 5.75e-05 bps in all three (< 0.01); B (replica CG−T) +1.9204 se 0.0445 on 7232768..7233279 (target 2.00 ± 0.15, passed); C (frozen port vs frozen Solidity) 6.03e-05 (j061), 5.99e-05 (j063), 5.65e-05 (j070) bps (< 0.1). Replica speeds: j061 918/s (6 cores), j063 635/s (4 cores), j070 677/s (12 procs).

## `best`

Highest holdout over_T among valid search rows: **j063** (2.5641 se 0.0136 vs j061 2.5585, j070 2.5359). Caveat: its real gain over T3 is only +0.0279 (se 0.0104) and +0.0640 (se 0.0138) per block, and its gain over the starting point on the holdout is +0.031 (se 0.0038) — not a path to +3.1 over HA. Gas: peaks 238,120 / 237,611 candidate afterSwap, 0 OOG. code_sha256 `7e588fca4d153358cdaa397eae44d9760d5192c222d83b36442aac037a95025b`.

```json
{"key":"j063","code_sha256":"7e588fca4d153358cdaa397eae44d9760d5192c222d83b36442aac037a95025b","code_zb64":"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","params":{"W1":[[-0.111595,0.644141,-0.463329,3.049371,-2.3485,0.614031,-1.429739,2.411745,-5.214866,-1.772879,-0.811424,-0.056791,0.502438,-2.101507,-0.496917,-0.323874,-1.203608,-0.077226,1.454431],[0.732427,0.736834,2.087008,2.703377,-0.101471,-0.215229,-0.467582,0.313124,-0.386013,0.159588,-0.126539,-1.774664,0.23016,-0.138552,-0.024881,1.100251,0.349409,0.124399,1.158901],[1.051804,3.78326,-1.179582,0.715082,0.924009,2.23424,-1.661978,1.010566,1.314979,-0.713929,0.399098,-0.09351,-0.366318,0.348469,0.224897,1.202414,-2.499175,-0.28726,-2.429604],[0.38452,-1.404438,0.922609,-0.370682,-0.957722,-0.33426,0.016131,-0.015687,0.831624,0.125793,0.165268,-1.83156,-0.35294,0.28711,0.199347,-1.099888,1.00916,-0.496926,-0.847081],[1.323043,0.397339,4.836468,-0.047604,-1.998355,-1.270921,-1.087473,-0.65234,-0.350013,0.521726,-0.021931,-2.296765,-0.053482,-1.085188,0.309048,-0.767309,2.205118,-0.087958,0.468388],[1.339988,-1.905912,0.346843,-1.406968,3.676618,1.410899,3.7838,0.571098,-0.508065,-0.032278,0.006357,3.740598,0.497151,-1.131331,0.206878,0.531458,-1.798853,-0.229719,-1.558828],[-0.943253,1.134171,-0.075669,-0.624544,2.455159,0.710052,-1.096756,2.314309,1.230845,-1.521933,1.167879,1.097439,1.42372,-1.094475,-1.147043,-1.463806,0.987684,-0.884491,2.557737],[-2.544712,-2.681709,3.316953,3.568031,-0.279308,-0.905134,-1.821735,2.548622,0.554745,-0.809915,2.431603,-2.122852,1.331394,0.20222,0.784222,1.002743,-0.481057,-1.80917,1.293379]],"b1":[1.695634,-1.605072,-4.266008,-0.973897,-1.357905,-2.059067,-0.518049,0.58676],"W2":[[0.243586,-2.053696,0.936671,-0.741546,-3.064935,-0.767388,-0.966841,0.828145],[0.338483,-1.909069,-1.679954,3.632106,0.243072,-2.118668,0.780564,-0.20256]],"b2":[1.390107,3.22796],"constants":{"ELAPSED_CAP":7,"SIGNAL_THRESHOLD":0.001213858145604775,"DIR_DECAY":0.7351807792152392,"ACT_DECAY":0.37089450149377673,"SIZE_DECAY":0.47714995979273045,"TOX_DECAY":0.9278704386636349,"SIGMA_DECAY":0.7230066679362842,"LAMBDA_DECAY":0.9917809097204768,"SIZE_BLEND_DECAY":0.8405119082815251,"TOX_BLEND_DECAY":0.23842849615367123,"ACT_BLEND_DECAY":0.9870184428349312,"PHAT_ALPHA":0.2544547047487652,"PHAT_ALPHA_RETAIL":0.055357411423175705,"PHAT_ALPHA_SAME":0.14174550462992616,"DIR_IMPACT_MULT":4,"GATE_SIGMA_MULT":6.657655984701647,"MIN_GATE":0.02989750505937308,"TRADE_TOX_BOOST":0.30562790751932184,"STALE_ATTRACT_FRAC":0.681292357353116,"RET_CAP":0.02275494186412903,"TOX_CAP":0.20634022455320594,"TRADE_RATIO_CAP":0.18654293223371024,"LAMBDA_CAP":1.2079055468962152,"STEP_COUNT_CAP":64,"BASE_FEE":0.000395988131048762,"SIGMA_COEF":0.20390636011356356,"LAMBDA_COEF":0.001284969398149582,"FLOW_SIZE_COEF":0.48578699093547195,"TOX_COEF":0.029470578284917077,"TOX_QUAD_COEF":1.5885181038698715,"ACT_COEF":8.754891170020448,"DIR_COEF":0.001764857258589,"DIR_TOX_COEF":0.007540389825676205,"STALE_DIR_COEF":0.7456772580980201,"SIGMA_TOX_COEF":0.060178960811209894,"TAIL_KNEE":0.05417286481065495,"TAIL_SLOPE_PROTECT":0.8453814301891062,"TAIL_SLOPE_ATTRACT":0.8331996962063852,"ANCHOR_COUNT_CAP":728,"ANCHOR_LINEAR_K":0.6899130209203691,"ANCHOR_QUADRATIC_K":1.1444676672757794},"scaling":{"feature_order":["sigmaHat","lambdaHat","sizeHat","toxEma","actEma","direction","(spot-staleHat)/staleHat","signed_trade_ratio","arb_probability","first_in_step","uncertainty","min(silent,8)/8","anchor_mean","raw_size","step_trade_count","prev_bid_fee","prev_ask_fee","steps_since_arb_like_print","router_implied_vanilla_spot_dev"],"feature_multipliers":[1000.0,2.0,500.0,500.0,500.0,1.0,500.0,500.0,2.0,2.0,1000000.0,1.0,500.0,647.054817,1.395287,220.617911,228.148131,0.412565,274.335453],"feature_offsets":[-1.0,-1.0,-1.0,-1.0,-1.0,-1.0,0.0,0.0,-1.0,-1.0,-1.0,0.0,-1.0,-0.618293,-2.037262,-0.803057,-0.808615,-0.823811,0.001772],"weight_fixed_point_scale":1000000,"input_clip":[-4,4],"hidden_relu_clip":[0,4],"output_unit_fee":0.0001,"note":"unchanged from CG"}}}
```

## `real_only`

T3 minus HA, each alone against 30 bps Vanilla, 1,024 fresh seeds per item (n=1024; none overlapping).

```json
[
 {
  "key": "j062",
  "seeds": "8944000..8945023",
  "n": "1024",
  "mean": "1.40388",
  "se": "0.0298431",
  "sum": "1437.57",
  "sumsq": "2951.14",
  "wins": "978",
  "gas": {
   "T3_afterSwap_peak": "235556",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "231567",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j064",
  "seeds": "8952000..8953023",
  "n": "1024",
  "mean": "1.40074",
  "se": "0.0304546",
  "sum": "1434.35",
  "sumsq": "2980.74",
  "wins": "981",
  "gas": {
   "T3_afterSwap_peak": "237441",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "231437",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j065",
  "seeds": "8956000..8957023",
  "n": "1024",
  "mean": "1.41086",
  "se": "0.02914",
  "sum": "1444.72",
  "sumsq": "2927.81",
  "wins": "970",
  "gas": {
   "T3_afterSwap_peak": "237461",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "233512",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j066",
  "seeds": "8960000..8961023",
  "n": "1024",
  "mean": "1.37655",
  "se": "0.0296426",
  "sum": "1409.59",
  "sumsq": "2860.84",
  "wins": "983",
  "gas": {
   "T3_afterSwap_peak": "235468",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "231471",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j067",
  "seeds": "8964000..8965023",
  "n": "1024",
  "mean": "1.41557",
  "se": "0.0292715",
  "sum": "1449.55",
  "sumsq": "2949.51",
  "wins": "987",
  "gas": {
   "T3_afterSwap_peak": "237588",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "231543",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j068",
  "seeds": "8968000..8969023",
  "n": "1024",
  "mean": "1.42324",
  "se": "0.0280054",
  "sum": "1457.4",
  "sumsq": "2895.83",
  "wins": "990",
  "gas": {
   "T3_afterSwap_peak": "235544",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "231567",
   "HA_OOG": "0"
  }
 },
 {
  "key": "j069",
  "seeds": "8972000..8973023",
  "n": "1024",
  "mean": "1.467",
  "se": "0.0409698",
  "sum": "1502.21",
  "sumsq": "3962.09",
  "wins": "984",
  "gas": {
   "T3_afterSwap_peak": "238223",
   "T3_OOG": "0",
   "HA_afterSwap_peak": "233632",
   "HA_OOG": "0"
  }
 }
]
```

Pooled over the 7 items (computed here from sum and sumsq): n=7168, mean 1.4140, se 0.0118 (unweighted item means range 1.3765–1.4670). T3 is therefore about +1.41 over HA on the real engine, consistent across items. Caveats: j069 covers gas stats for only 768 of its markets (0 OOG on 8,358,924 afterSwap calls, per its note) and its se (0.041) is larger because of one outlier seed (8972082 at +31.0); j067's 1,024 seeds were completed in full; j064's and j066's notes record 0 failed calls. T3 afterSwap peaks 235,468–238,223 (≤ 245,000), HA peaks 231,437–233,632, 0 OOG in all.

Implication for the goal: T3 ≈ HA + 1.41; best search candidate ≈ T3 + 0.03 to + 0.06 → ≈ HA + 1.45 (different seeds per item, so indicative only), against the ≈ +3.1 needed. (Baseline note: the goal text puts HA at +1.53 over RS2; this merge did not measure RS2 or the board.)

## What failed or was not done

- 7 of 10 items (j062, j064, j065, j066, j067, j068, j069) ran no search because their replica speed measured 42–450 markets/s on their machines (j062 42, j064 125, j065 104, j066 450, j067 121, j068 101, j069 180) — below the 500 threshold; that is a result of their machine/load, not of the policy. The requested settings of those items (including j065 setting D, j067 setting E) were therefore never tested; no information on them exists here.
- Gate C was not run in real-only items. Retail/arb edge deltas are null everywhere (engine does not expose them).
- j070's first real block counts volume/pnl over 1,023 markets (one baseline record missed); edge stats use all 1,024.
- No item reached the top-10 target; no leaderboard placement is claimed by any item and none is claimed here.