{
 "sources": [
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/README.md",
   "quote": "**What separates +4.26 from #1 is one call:** is the first trade of a step the arb or retail. The filter is wrong 16% of the time. With the true answer the same code scores +6.98."
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/README.md",
   "quote": "10000  448.261189  443.455598  443.522608  443.663520"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/README.md",
   "quote": "buying more pins with tighter quotes (kappa above 1e-5 costs more than it returns)"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/README.md",
   "quote": "The file was built with an information weight (kappa = 1e-5, the third argument of `qmdp.py`) that favors quotes the arb will hit, so even level 0 differs from `pol_mid.json`, in about a quarter of the cells; built with kappa 0, level 0 equals it."
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/README.md",
   "quote": "DP_SAVEVN=1 DP_AMINL=-12 DP_NA=43 python dp.py solve 0.8 9.45 20 pol_mid.json     # 536.821, tables identical to the file here"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/qmdp.py",
   "quote": "(a, b)* = argmax  sum_e  bel(e) * V_N[phase](w - e, w2(e), a - e, b + e)"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/qmdp.py",
   "quote": "expected drop in belief variance (bps^2) over the next step boundary, with the arb/retail ambiguity:"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/qmdp.py",
   "quote": "IG = igtab2(s, T['sigma']) if os.environ.get('IG2') else igtab(s, T['sigma'])"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/real.py",
   "quote": "if o.get('oraclearb', 0): pi = 1.0 if arb else 0.0"
  },
  {
   "url": "https://raw.githubusercontent.com/gtmcknight/amm-sim/4812a6382311624e1506aca131ca35c378f930d9/research/solved/gates.py",
   "quote": "c = R.run(seed, real.RealPolicy(T, {'soft': 1.0, 'qm': QM}))['edge']"
  },
  {
   "url": "https://api.imd.fun/artifacts/8a6e303b662e1a5ea9bb47f31acedd3988ef65b3ddb66d074517aa59ad41ca40",
   "quote": "## Python reference (complete; `qmref.py`, stdlib only, needs the 10,167-byte QMS1 blob from qm-spline's artifact)"
  },
  {
   "url": "https://api.imd.fun/artifacts/8a6e303b662e1a5ea9bb47f31acedd3988ef65b3ddb66d074517aa59ad41ca40",
   "quote": "The source is 42,990 bytes, sha256 **`c68887cff61180e12ba2b1ceb333edf76a830f91f63b5d775a8806fb9ec08aca`**."
  },
  {
   "url": "https://api.imd.fun/artifacts/d360bea6c8c982717b7b43a8069a87013169c89ee3e4da25622db4d880b936d4",
   "quote": "| report 10000..12047 | COMBO-qt1 | 2048 | +0.597391 | 0.013388 | 44.6 | 1753/2048 |"
  },
  {
   "url": "https://api.imd.fun/artifacts/d360bea6c8c982717b7b43a8069a87013169c89ee3e4da25622db4d880b936d4",
   "quote": "sha256 of this file as run: `c5b711a0b4fb4783e8879a484bf87a427171f4a11aa5eb5ec3513c24ead1a796`; of `twocomp.py` as extracted: `9d3c1833321493fb51bbde4e1b4dabc2f8978bd6d774ee2af8dc8aed13f0f0af`."
  },
  {
   "url": "https://api.imd.fun/artifacts/d360bea6c8c982717b7b43a8069a87013169c89ee3e4da25622db4d880b936d4",
   "quote": "| 10000 | 443.101741 | 443.690614 | 444.195330 |"
  },
  {
   "url": "https://api.imd.fun/artifacts/d3c849db2ac31e567b57e52b80b86e8c24c842e5b4125b9d75c78e3cb07115b9",
   "quote": "The first own trade in a step produces an arb point mass `(pi, lp, 0)` and a retail truncated Gaussian `(1-pi, mt, vt)`."
  },
  {
   "url": "https://api.imd.fun/artifacts/17d249ab0c33fc44931d9b105cf3f9e3846a560bd518b67310c2a56defacdecc",
   "quote": "On the holdout the paired loss against the full qm tables is **0.0632, se 0.0072, n = 2048**."
  },
  {
   "url": "https://api.imd.fun/artifacts/17d249ab0c33fc44931d9b105cf3f9e3846a560bd518b67310c2a56defacdecc",
   "quote": "It is 10,167 bytes, sha256 `3ac7cb8811ecfb6a17f499aa4e9b70dd55f7fb6432b3aec3b904e11a3dece38e`."
  }
 ]
}