{
  "series": [
    {
      "name": "P4 / memchain-to-trace",
      "points": [
        {
          "date": "2026-07-02",
          "k": -0.0017,
          "U": 4096,
          "payload": 4096,
          "r": 1.0,
          "fixed": {
            "binary": 16984,
            "header": 12,
            "total": 16996
          },
          "S_over_U_enwik9": 1.000016996,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run memchain-to-trace --ofra tests/pprog/p4-memchain-to-trace.md 8192 --compress <model>  < <first 4096 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 4096,
            "cmpr_checksum": "BB9517998A955A5F"
          },
          "program": {
            "ofra": "tests/pprog/p4-memchain-to-trace.md",
            "ofra_cmpr_checksum": "C9A68986838DA376",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_main": "D1BB22E0F5E0DFE1"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 1.000016996 ; k where S/U = .99^k : k = -0.0017",
          "repro": {
            "source": "pprog/p4-memchain-to-trace.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 4096 enwik9 > sample\ncmpr/dist/cmpr --run memchain-to-trace --ofra tests/pprog/p4-memchain-to-trace.md 8192 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run memchain-to-trace --ofra tests/pprog/p4-memchain-to-trace.md --decompress model | cmp - sample && echo round-trip OK"
          }
        }
      ]
    },
    {
      "name": "P7 / order-1-markov",
      "points": [
        {
          "date": "2026-07-05",
          "k": 41.9448,
          "U": 4096,
          "payload": 2687,
          "r": 0.656005859375,
          "fixed": {
            "binary": 16984,
            "header": 20,
            "markov table": 256,
            "total": 17260
          },
          "S_over_U_enwik9": 0.656023119375,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run order-1-markov --ofra tests/pprog/p7-hutter-order-1-markov.md 8192 --compress <model>  < <first 4096 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 4096,
            "cmpr_checksum": "BB9517998A955A5F"
          },
          "program": {
            "ofra": "tests/pprog/p7-hutter-order-1-markov.md",
            "ofra_cmpr_checksum": "95A52EB67BE2FDE1",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_main": "4B8DD41BE7192684"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.656023119 ; k where S/U = .99^k : k = 41.9448",
          "repro": {
            "source": "pprog/p7-hutter-order-1-markov.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 4096 enwik9 > sample\ncmpr/dist/cmpr --run order-1-markov --ofra tests/pprog/p7-hutter-order-1-markov.md 8192 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run order-1-markov --ofra tests/pprog/p7-hutter-order-1-markov.md --decompress model | cmp - sample && echo round-trip OK"
          }
        }
      ]
    },
    {
      "name": "P8 / words",
      "points": [
        {
          "date": "2026-07-06",
          "k": 20.9567,
          "U": 4096,
          "payload": 3318,
          "r": 0.81005859375,
          "fixed": {
            "binary": 21120,
            "header": 32,
            "markov table": 256,
            "total": 21408
          },
          "S_over_U_enwik9": 0.81008000175,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 8192 --compress <model>  < <first 4096 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 4096,
            "cmpr_checksum": "BB9517998A955A5F"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_main": "C25A819EFE821184"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.810080002 ; k where S/U = .99^k : k = 20.9567",
          "repro": {
            "source": "pprog/p8-words.md",
            "report": "pprog/p8-report.pdf",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 4096 enwik9 > sample\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 8192 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md --decompress model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 7.1542,
          "U": 10000,
          "payload": 9306,
          "r": 0.9306,
          "fixed": {
            "binary": 21248,
            "header": 32,
            "markov table": 256,
            "total": 21536
          },
          "S_over_U_enwik9": 0.930621536,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 10000 --compress <model>  < <first 10000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 10000,
            "cmpr_checksum": "1929DCD7668ED722"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_codec": "C03CBD89BB417580",
              "#pp_markov": "0461BE2400B9F5C9",
              "#pp_words": "0A48F971D6AD623C",
              "#pp_sections": "78EE30066FFBEDC5",
              "#pp_spellings": "F53F67A69990DA16",
              "#pp_compress": "C2114CFC3B003156",
              "#pp_decompress": "8E3F3C3116DC0DB5",
              "#pp_main": "309BF1EF7A101C7F"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.930621536 ; k where S/U = .99^k : k = 7.1542",
          "repro": {
            "source": "pprog/p8-words.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 10000 enwik9 > sample\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 10000 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md --decompress model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 28.8098,
          "U": 100000,
          "payload": 74858,
          "r": 0.74858,
          "fixed": {
            "binary": 21248,
            "header": 32,
            "markov table": 256,
            "total": 21536
          },
          "S_over_U_enwik9": 0.748601536,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 100000 --compress <model>  < <first 100000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 100000,
            "cmpr_checksum": "B455AFE32721B6ED"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_codec": "C03CBD89BB417580",
              "#pp_markov": "0461BE2400B9F5C9",
              "#pp_words": "0A48F971D6AD623C",
              "#pp_sections": "78EE30066FFBEDC5",
              "#pp_spellings": "F53F67A69990DA16",
              "#pp_compress": "C2114CFC3B003156",
              "#pp_decompress": "8E3F3C3116DC0DB5",
              "#pp_main": "309BF1EF7A101C7F"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.748601536 ; k where S/U = .99^k : k = 28.8098",
          "repro": {
            "source": "pprog/p8-words.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 100000 enwik9 > sample\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 100000 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md --decompress model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 43.2402,
          "U": 1000000,
          "payload": 647516,
          "r": 0.647516,
          "fixed": {
            "binary": 21248,
            "header": 32,
            "markov table": 256,
            "total": 21536
          },
          "S_over_U_enwik9": 0.647537536,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 1000000 --compress <model>  < <first 1000000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 1000000,
            "cmpr_checksum": "A3F49F596D6781C4"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_codec": "C03CBD89BB417580",
              "#pp_markov": "0461BE2400B9F5C9",
              "#pp_words": "0A48F971D6AD623C",
              "#pp_sections": "78EE30066FFBEDC5",
              "#pp_spellings": "F53F67A69990DA16",
              "#pp_compress": "C2114CFC3B003156",
              "#pp_decompress": "8E3F3C3116DC0DB5",
              "#pp_main": "309BF1EF7A101C7F"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.647537536 ; k where S/U = .99^k : k = 43.2402",
          "repro": {
            "source": "pprog/p8-words.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 1000000 enwik9 > sample\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 1000000 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md --decompress model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 51.5588,
          "U": 10000000,
          "payload": 5955799,
          "r": 0.5955799,
          "fixed": {
            "binary": 21248,
            "header": 32,
            "markov table": 256,
            "total": 21536
          },
          "S_over_U_enwik9": 0.595601436,
          "round_trip": "verified",
          "command": "/home/admin/cmpr-src/cmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 10000000 --compress <model>  < <first 10000000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 10000000,
            "cmpr_checksum": "A71F392E34B6DCF8"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {
              "#pp_rt": "5E1BEF6908FAFFFC",
              "#pp_es": "C0EE79AF21BEF489",
              "#pp_pattern": "8E45CE96A255A69D",
              "#pp_codec": "C03CBD89BB417580",
              "#pp_markov": "0461BE2400B9F5C9",
              "#pp_words": "0A48F971D6AD623C",
              "#pp_sections": "78EE30066FFBEDC5",
              "#pp_spellings": "F53F67A69990DA16",
              "#pp_compress": "C2114CFC3B003156",
              "#pp_decompress": "8E3F3C3116DC0DB5",
              "#pp_main": "309BF1EF7A101C7F"
            }
          },
          "raw": "run-ofra: hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.595601436 ; k where S/U = .99^k : k = 51.5588",
          "repro": {
            "source": "pprog/p8-words.md",
            "sh": "# from a cmpr-src checkout (build: make; binary at cmpr/dist/cmpr);\n# enwik9 from http://mattmahoney.net/dc/enwik9.zip\nhead -c 10000000 enwik9 > sample\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md 10000000 --compress model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\ncmpr/dist/cmpr --run words --ofra tests/pprog/p8-words.md --decompress model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 60.391,
          "U": 100000000,
          "payload": 54498896,
          "r": 0.54498896,
          "fixed": {
            "binary": 21600,
            "header": 32,
            "markov table": 256,
            "total": 21888
          },
          "S_over_U_enwik9": 0.545010848,
          "round_trip": "verified",
          "command": "p8-opt 100000000 1000000 c <model>  < <first 100000000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 100000000,
            "cmpr_checksum": "08498AD2DD35973F"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {},
            "runner": "tests/pprog/p8-opt.c",
            "runner_cmpr_checksum": "2779F989DFE3BC65",
            "equivalence": "model files byte-identical to the ofra-generated reference (verified at 10^4 and 10^5, 2026-07-08)"
          },
          "raw": "hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.545010848 ; k where S/U = .99^k : k = 60.3910",
          "repro": {
            "source": "pprog/p8-words.md",
            "runner_source": "pprog/p8-opt.c",
            "sh": "# optimized standalone P8 (model byte-identical to the ofra-generated\n# reference); from a cmpr-src checkout; enwik9 from mattmahoney.net\ncc -O2 -o p8-opt tests/pprog/p8-opt.c -lm\nhead -c 100000000 enwik9 > sample\n./p8-opt 100000000 1000000 c model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\n./p8-opt 100000000 1000000 d model | cmp - sample && echo round-trip OK"
          }
        },
        {
          "date": "2026-07-06",
          "k": 58.0479,
          "U": 1000000000,
          "payload": 557975528,
          "r": 0.557975528,
          "fixed": {
            "binary": 21600,
            "header": 32,
            "markov table": 256,
            "total": 21888
          },
          "S_over_U_enwik9": 0.5579974160000001,
          "round_trip": "verified",
          "command": "p8-opt 1000000000 1000000 c <model>  < <first 1000000000 bytes of enwik9>",
          "input": {
            "desc": "enwik9-prefix",
            "bytes": 1000000000,
            "cmpr_checksum": "13EE236047DF3050"
          },
          "program": {
            "ofra": "tests/pprog/p8-words.md",
            "ofra_cmpr_checksum": "7D265165E185573E",
            "blocks": {},
            "runner": "tests/pprog/p8-opt.c",
            "runner_cmpr_checksum": "2779F989DFE3BC65",
            "equivalence": "model files byte-identical to the ofra-generated reference (verified at 10^4 and 10^5, 2026-07-08)"
          },
          "raw": "hutter: extrapolated whole-enwik9 S/U = r + fixed/1e9 = 0.557997416 ; k where S/U = .99^k : k = 58.0479",
          "repro": {
            "source": "pprog/p8-words.md",
            "runner_source": "pprog/p8-opt.c",
            "sh": "# optimized standalone P8 (model byte-identical to the ofra-generated\n# reference); from a cmpr-src checkout; enwik9 from mattmahoney.net\ncc -O2 -o p8-opt tests/pprog/p8-opt.c -lm\nhead -c 1000000000 enwik9 > sample\n./p8-opt 1000000000 1000000 c model < sample\n#   report on stderr: U, archive P, r = P/U, fixed cost, extrapolated k\n./p8-opt 1000000000 1000000 d model | cmp - sample && echo round-trip OK"
          }
        }
      ]
    }
  ],
  "metric": {
    "definition": "S/U = 0.99^k ;  k = log(S/U) / log(0.99)",
    "U_enwik9": 1000000000,
    "note": "k extrapolated to enwik9; +1 k == 1% smaller archive. Extrapolation holds the small-sample rate r constant, so it is a conservative floor for models that learn more from more data.",
    "generated": "2026-07-08T06:26:50+00:00"
  },
  "reference_lines": [
    {
      "label": "no compression",
      "S": 1000000000,
      "emphasis": "zero",
      "k": -0.0
    },
    {
      "label": "enwik9.zip (dumb baseline)",
      "S": 322592222,
      "k": 112.57
    },
    {
      "label": "Hutter record  L",
      "S": 110793128,
      "k": 218.9072
    },
    {
      "label": "prize: 1% claim (0.99\u00b7L)",
      "S": 109685197,
      "k": 219.9072
    }
  ],
  "predictions": [
    {
      "name": "P8 component-model extrapolation",
      "short": "P8 pred",
      "source": "#pprog_p8_words_20260706",
      "date": "2026-07-06",
      "input_rate_Bps": 200000,
      "days_per_decade": 0.6,
      "ladder": [
        {
          "scale": "10^4",
          "M": 10000,
          "r": 0.931,
          "kind": "measured",
          "k": 7.1,
          "train_s": 0.05,
          "offset_days": 0.0
        },
        {
          "scale": "10^5",
          "M": 100000,
          "r": 0.749,
          "kind": "measured",
          "k": 28.8,
          "train_s": 0.5,
          "offset_days": 0.6
        },
        {
          "scale": "10^6",
          "M": 1000000,
          "r": 0.64,
          "kind": "extrapolated",
          "k": 44.4,
          "train_s": 5.0,
          "offset_days": 1.2,
          "ci95": {
            "r_lo": 0.55,
            "r_hi": 0.73,
            "k_lo": 31.3,
            "k_hi": 59.5
          }
        },
        {
          "scale": "10^7",
          "M": 10000000,
          "r": 0.57,
          "kind": "extrapolated",
          "k": 55.9,
          "train_s": 50.0,
          "offset_days": 1.8,
          "ci95": {
            "r_lo": 0.48,
            "r_hi": 0.66,
            "k_lo": 41.3,
            "k_hi": 73.0
          }
        },
        {
          "scale": "10^8",
          "M": 100000000,
          "r": 0.55,
          "kind": "extrapolated",
          "k": 59.5,
          "train_s": 500.0,
          "offset_days": 2.4,
          "ci95": {
            "r_lo": 0.46,
            "r_hi": 0.64,
            "k_lo": 44.4,
            "k_hi": 77.3
          }
        },
        {
          "scale": "10^9",
          "M": 1000000000,
          "r": 0.54,
          "kind": "extrapolated",
          "k": 61.3,
          "train_s": 5000.0,
          "offset_days": 3.0,
          "ci95": {
            "r_lo": 0.45,
            "r_hi": 0.63,
            "k_lo": 46.0,
            "k_hi": 79.5
          }
        }
      ],
      "sigma_r": 0.045
    }
  ]
}