Datasets:
Tasks:
Tabular Classification
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
economics
quantitative-finance
causal-inference
macroeconomics
housing-economics
market-microstructure
License:
File size: 62,052 Bytes
ebcde1f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 | """Frozen public aggregate-trade research-run producer.
This producer is deliberately narrower than the synthetic vertical slice. It
loads one explicitly manifested, bounded public trade ingestion; proves trade-ID
and availability-clock continuity before deriving research epochs; evaluates
each instrument independently; and publishes predictive diagnostics only. It
never invokes the execution simulator or calculates P&L.
"""
from __future__ import annotations
import hashlib
import json
import math
import os
import tempfile
from collections.abc import Mapping, Sequence
from dataclasses import asdict, dataclass
from datetime import UTC, datetime
from pathlib import Path
from typing import Any, cast
import numpy as np
import polars as pl
from microstructure.config import ProjectConfig
from microstructure.provenance import (
provenance_header,
read_json,
sha256_file,
write_json,
)
from microstructure.public_data import PublicTrades, read_public_trades
from microstructure.reporting import (
load_run_bundle,
render_executive_memo,
render_model_comparison,
render_technical_report,
write_checksum_manifest,
)
from microstructure.research.analysis import (
feature_stability_summary,
ofi_future_return_association,
)
from microstructure.research.models import (
block_bootstrap_metric,
evaluate_model_ladder,
paired_block_bootstrap_difference,
)
from microstructure.research.splits import WalkForwardPlan, expanding_walk_forward_splits
from microstructure.research.trade_only import (
build_trade_only_research_frame,
validate_trade_only_temporal_contract,
)
PUBLIC_PIPELINE_SCHEMA_VERSION = "1.0.0"
PUBLIC_EVIDENCE_TIER = "PUBLIC_SAMPLE_PARTIAL"
EXECUTION_EXCLUSION_REASON = (
"Not run: aggregate-trade history has no contemporaneous quotes, depth, queue state, "
"or local receipt clock, so execution, fees-to-alpha conversion, P&L, fills, and capacity "
"would be unsupported claims."
)
HYPOTHESIS_BASELINE = "historical_prior"
HYPOTHESIS_BLOCK_POLICY = "fixed_contiguous_2x_label_horizon"
HYPOTHESIS_CAVEAT = (
"Exploratory paired percentile interval on a bounded retrospective sample; it is a "
"dependence diagnostic, not a p-value, confirmatory significance test, or basis for "
"rejecting H0."
)
CROSS_INSTRUMENT_CONCLUSION = (
"No pooled or cross-instrument conclusion is inferred: BTCUSDT and ETHUSDT are separate "
"sample-specific diagnostics. They cannot support a persistent-alpha claim, regardless "
"of their point-estimate directions."
)
__all__ = [
"EXECUTION_EXCLUSION_REASON",
"PUBLIC_EVIDENCE_TIER",
"PUBLIC_PIPELINE_SCHEMA_VERSION",
"PublicPipelineError",
"produce_public_trade_run",
]
class PublicPipelineError(RuntimeError):
"""Raised when the frozen public run cannot be produced honestly."""
@dataclass(frozen=True, slots=True)
class _SymbolResult:
symbol: str
research: pl.DataFrame
evaluation: pl.DataFrame
plan: WalkForwardPlan
predictions: pl.DataFrame
comparison: pl.DataFrame
selected_predictions: pl.DataFrame
selected_model: str
feature_columns: tuple[str, ...]
temporal_audit: Mapping[str, Any]
hypothesis_evaluation: Mapping[str, Any]
def _json_safe(value: Any) -> Any:
if isinstance(value, np.generic):
return _json_safe(value.item())
if isinstance(value, float) and not math.isfinite(value):
return None
if isinstance(value, Path):
return str(value)
if isinstance(value, Mapping):
return {str(key): _json_safe(item) for key, item in value.items()}
if isinstance(value, (list, tuple)):
return [_json_safe(item) for item in value]
return value
def _write_json(path: Path, payload: Mapping[str, Any] | list[Any]) -> None:
clean = _json_safe(payload)
if not isinstance(clean, (dict, list)):
raise TypeError("JSON artifact payload must be an object or list")
write_json(path, clean)
def _atomic_write_text(path: Path, content: str) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
descriptor, temporary_name = tempfile.mkstemp(
dir=path.parent,
prefix=f".{path.name}.",
suffix=".tmp",
text=True,
)
try:
with os.fdopen(descriptor, "w", encoding="utf-8", newline="\n") as handle:
handle.write(content.rstrip() + "\n")
os.replace(temporary_name, path)
except BaseException:
Path(temporary_name).unlink(missing_ok=True)
raise
def _freeze_protocol(config: ProjectConfig, stage: Path) -> tuple[str, str]:
source = config.project_root / "docs" / "PUBLIC_TRADE_PROTOCOL.md"
if not source.is_file():
raise PublicPipelineError(f"frozen public trade protocol is missing: {source}")
before = sha256_file(source)
try:
content = source.read_bytes()
except OSError as exc:
raise PublicPipelineError(f"cannot read frozen public trade protocol: {source}") from exc
after = sha256_file(source)
if before != after or hashlib.sha256(content).hexdigest() != before:
raise PublicPipelineError("public trade protocol changed while it was being frozen")
relative = "protocol/PUBLIC_TRADE_PROTOCOL.md"
destination = stage / relative
destination.parent.mkdir(parents=True, exist_ok=True)
descriptor, temporary_name = tempfile.mkstemp(
dir=destination.parent,
prefix=f".{destination.name}.",
suffix=".tmp",
)
try:
with os.fdopen(descriptor, "wb") as handle:
handle.write(content)
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary_name, destination)
except BaseException:
Path(temporary_name).unlink(missing_ok=True)
raise
if sha256_file(destination) != before:
raise PublicPipelineError("frozen protocol copy does not match its source SHA-256")
return relative, before
def _utc_from_ns(timestamp_ns: int) -> str:
seconds, nanoseconds = divmod(timestamp_ns, 1_000_000_000)
instant = datetime.fromtimestamp(seconds, tz=UTC)
return f"{instant:%Y-%m-%dT%H:%M:%S}.{nanoseconds:09d}Z"
def _stable_sha256(payload: Mapping[str, Any]) -> str:
encoded = json.dumps(payload, sort_keys=True, separators=(",", ":"), allow_nan=False)
return hashlib.sha256(encoded.encode("utf-8")).hexdigest()
def _mapping(value: object) -> Mapping[str, Any] | None:
if not isinstance(value, Mapping) or not all(isinstance(key, str) for key in value):
return None
return cast(Mapping[str, Any], value)
def _project_path(path: Path, project_root: Path) -> str:
resolved = path.resolve()
try:
return resolved.relative_to(project_root.resolve()).as_posix()
except ValueError:
return str(resolved)
def _verified_manifest_object(path: Path, expected_sha256: str) -> Mapping[str, Any]:
before = sha256_file(path)
if before != expected_sha256:
raise PublicPipelineError(f"input manifest changed after verified loading: {path}")
payload = _mapping(read_json(path))
after = sha256_file(path)
if after != expected_sha256 or after != before:
raise PublicPipelineError(f"input manifest changed while capturing lineage: {path}")
if payload is None:
raise PublicPipelineError(f"verified input manifest is not a JSON object: {path}")
return payload
def _input_lineage(public: PublicTrades, config: ProjectConfig) -> dict[str, Any]:
ingestion = _verified_manifest_object(
public.ingestion_manifest_path, public.ingestion_manifest_sha256
)
dataset = _verified_manifest_object(
public.dataset_manifest_path, public.dataset_manifest_sha256
)
manifest_hashes = set(public.input_manifest_sha256s)
data_hashes = set(public.raw_artifact_sha256s)
raw_entries = ingestion.get("raw_artifacts")
if not isinstance(raw_entries, list):
raise PublicPipelineError("verified ingestion manifest has no raw artifact array")
for entry_value in raw_entries:
entry = _mapping(entry_value)
if entry is None:
raise PublicPipelineError("verified raw artifact entry is not an object")
raw_sha = entry.get("sha256")
raw_manifest_sha = entry.get("manifest_sha256")
if isinstance(raw_sha, str):
data_hashes.add(raw_sha)
raw_path = public.ingestion_manifest_path.parent.parent / str(entry.get("path"))
if sha256_file(raw_path) != raw_sha:
raise PublicPipelineError(f"raw input changed while capturing lineage: {raw_path}")
if isinstance(raw_manifest_sha, str):
manifest_hashes.add(raw_manifest_sha)
raw_manifest_path = public.ingestion_manifest_path.parent.parent / str(
entry.get("manifest_path")
)
if sha256_file(raw_manifest_path) != raw_manifest_sha:
raise PublicPipelineError(
f"raw manifest changed while capturing lineage: {raw_manifest_path}"
)
part_entries = dataset.get("artifacts")
if not isinstance(part_entries, list):
raise PublicPipelineError("verified dataset manifest has no artifact array")
for entry_value in part_entries:
entry = _mapping(entry_value)
if entry is None:
raise PublicPipelineError("verified normalized artifact entry is not an object")
data_sha = entry.get("data_sha256")
part_manifest_sha = entry.get("manifest_sha256")
if isinstance(data_sha, str):
data_hashes.add(data_sha)
data_path = public.dataset_manifest_path.parent.parent / str(entry.get("data_path"))
if sha256_file(data_path) != data_sha:
raise PublicPipelineError(
f"normalized part changed while capturing lineage: {data_path}"
)
if isinstance(part_manifest_sha, str):
manifest_hashes.add(part_manifest_sha)
part_manifest_path = public.dataset_manifest_path.parent.parent / str(
entry.get("manifest_path")
)
if sha256_file(part_manifest_path) != part_manifest_sha:
raise PublicPipelineError(
"normalized part manifest changed while capturing lineage: "
f"{part_manifest_path}"
)
return {
"ingestion_manifest": {
"absolute_path": str(public.ingestion_manifest_path.resolve()),
"project_relative_or_absolute_path": _project_path(
public.ingestion_manifest_path, config.project_root
),
"sha256": public.ingestion_manifest_sha256,
},
"normalized_dataset_manifest": {
"absolute_path": str(public.dataset_manifest_path.resolve()),
"project_relative_or_absolute_path": _project_path(
public.dataset_manifest_path, config.project_root
),
"sha256": public.dataset_manifest_sha256,
},
"normalized_parts": [
_project_path(path, config.project_root) for path in public.part_paths
],
"raw_artifacts": [
_project_path(path, config.project_root) for path in public.raw_artifact_paths
],
"raw_manifests": [
_project_path(path, config.project_root) for path in public.raw_manifest_paths
],
"manifest_sha256": sorted(manifest_hashes),
"data_sha256": sorted(data_hashes),
}
def _derive_verified_continuity(
public: PublicTrades, config: ProjectConfig
) -> tuple[pl.DataFrame, list[dict[str, Any]]]:
if public.rows > public.row_bound:
raise PublicPipelineError("materialized public rows exceed the verified configured bound")
raw = public.polars_trades
expected_symbols = set(config.data.symbols)
if set(str(value) for value in raw.get_column("symbol").unique()) != expected_symbols:
raise PublicPipelineError("materialized public symbols do not match the configuration")
if raw.get_column("continuity_id").null_count() != raw.height:
raise PublicPipelineError(
"archive normalized rows must retain null source continuity before derivation"
)
derived_frames: list[pl.DataFrame] = []
audits: list[dict[str, Any]] = []
for symbol in config.data.symbols:
source = raw.filter(pl.col("symbol") == symbol).sort("trade_id")
if source.is_empty():
raise PublicPipelineError(f"verified public input has no rows for {symbol}")
ids = source.get_column("trade_id").to_numpy().astype(np.int64)
available = source.get_column("available_ts_ns").to_numpy().astype(np.int64)
event = source.get_column("event_ts_ns").to_numpy().astype(np.int64)
if np.any(np.diff(ids) != 1):
raise PublicPipelineError(f"aggregate trade IDs are not contiguous by one for {symbol}")
if np.any(np.diff(available) < 0):
raise PublicPipelineError(
f"availability clock reverses in aggregate-trade-ID order for {symbol}"
)
if np.any(available < event):
raise PublicPipelineError(f"availability precedes event time for {symbol}")
bases = set(str(value) for value in source.get_column("availability_basis").unique())
if bases != {"exchange_event_time_proxy"}:
raise PublicPipelineError(
f"historical public trades require the exchange-event-time proxy for {symbol}"
)
if source.get_column("received_ts_ns").null_count() != source.height:
raise PublicPipelineError(
f"historical public trades cannot claim local receipt time for {symbol}"
)
continuity_id = (
f"public-aggtrade:{public.ingestion_manifest_sha256[:16]}:{symbol}:"
f"{int(ids[0])}-{int(ids[-1])}"
)
derived_frames.append(source.with_columns(pl.lit(continuity_id).alias("continuity_id")))
audits.append(
{
"symbol": symbol,
"rows": source.height,
"first_trade_id": int(ids[0]),
"last_trade_id": int(ids[-1]),
"trade_id_step": 1,
"trade_ids_contiguous": True,
"availability_clock_nondecreasing": True,
"tied_availability_rows": int(np.count_nonzero(np.diff(available) == 0)),
"availability_basis": "exchange_event_time_proxy",
"local_receipt_time_available": False,
"derived_continuity_id": continuity_id,
"derivation_timing": "assigned only after ID and clock verification",
"source_rows_mutated": False,
}
)
return pl.concat(derived_frames).sort(["symbol", "trade_id"]), audits
def _trade_feature_columns(config: ProjectConfig, frame: pl.DataFrame) -> tuple[str, ...]:
columns: list[str] = ["log_trade_return_1"]
for window in config.features.trade_windows:
columns.extend(
[
f"signed_trade_volume_w{window}",
f"trade_volume_w{window}",
f"trade_imbalance_w{window}",
]
)
intensity = config.features.intensity_window
columns.extend([f"trade_count_w{intensity}", f"trade_intensity_w{intensity}"])
columns.append(f"realized_volatility_w{config.features.volatility_window}")
selected = tuple(dict.fromkeys(columns))
missing = sorted(set(selected).difference(frame.columns))
if missing:
raise PublicPipelineError(f"declared trade-only model features are missing: {missing}")
return selected
def _serialize_plan(plan: WalkForwardPlan) -> dict[str, Any]:
return {
"contract": (
"single-symbol decision-time buckets; expanding training; feature-ready and "
"uncensored evaluation; labels ending at or beyond evaluation are purged; "
"configured embargo applied"
),
"index_basis": "zero-based row positions in this symbol's evaluation_frame.parquet",
"decision_time_count": plan.decision_time_count,
"folds": [
{
"fold_id": fold.fold_id,
"train_indices": [int(value) for value in fold.train_indices.tolist()],
"validation_indices": [int(value) for value in fold.validation_indices.tolist()],
"train_start_ts_ns": fold.train_start_ts_ns,
"train_end_ts_ns": fold.train_end_ts_ns,
"validation_start_ts_ns": fold.validation_start_ts_ns,
"validation_end_ts_ns": fold.validation_end_ts_ns,
"purged_rows": fold.purged_rows,
"embargoed_time_buckets": fold.embargoed_time_buckets,
}
for fold in plan.folds
],
"final_train_indices": [int(value) for value in plan.final_train_indices.tolist()],
"test_indices": [int(value) for value in plan.test_indices.tolist()],
"test_start_ts_ns": plan.test_start_ts_ns,
"test_end_ts_ns": plan.test_end_ts_ns,
"test_used_for_selection": False,
}
def _add_fixed_blocks(predictions: pl.DataFrame, *, block_width: int) -> pl.DataFrame:
group = ["split", "fold_id"]
return (
predictions.with_columns(
pl.col("decision_sequence").min().over(group).alias("_block_origin_trade_id")
)
.with_columns(
((pl.col("decision_sequence") - pl.col("_block_origin_trade_id")) // block_width)
.cast(pl.Int64)
.alias("_block_index")
)
.with_columns(
pl.concat_str(
[
"symbol",
"split",
pl.col("fold_id").cast(pl.String),
pl.col("_block_index").cast(pl.String),
],
separator=":",
).alias("bootstrap_block"),
(pl.col("_block_origin_trade_id") + pl.col("_block_index") * block_width).alias(
"bootstrap_block_start_trade_id"
),
)
.with_columns(
(pl.col("bootstrap_block_start_trade_id") + block_width - 1).alias(
"bootstrap_block_end_trade_id"
),
pl.lit(block_width, dtype=pl.Int64).alias("bootstrap_block_width_trades"),
pl.lit("fixed_contiguous_2x_label_horizon").alias("bootstrap_block_policy"),
)
.drop("_block_origin_trade_id", "_block_index")
)
def _bootstrap_comparison(
comparison: pl.DataFrame,
predictions: pl.DataFrame,
*,
symbol: str,
selected_model: str,
metric: str,
n_bootstrap: int,
seed: int,
horizon: int,
) -> pl.DataFrame:
rows: list[dict[str, Any]] = []
block_width = 2 * horizon
for index, source_row in enumerate(
comparison.sort(["split", "fold_id", "requested_model"]).to_dicts()
):
row = dict(source_row)
requested_model = str(row.get("requested_model", row["model"]))
row["selected_on_validation"] = requested_model == selected_model
row["selected_on"] = "validation" if requested_model == selected_model else None
row["selection_contract"] = "validation folds only; final test opened after selection"
row["test_used_for_selection"] = False
row["evidence_tier"] = PUBLIC_EVIDENCE_TIER
row["execution_evaluated"] = False
row[f"{metric}_ci_low"] = None
row[f"{metric}_ci_high"] = None
row["bootstrap_status"] = None
row["bootstrap_blocks"] = None
row["bootstrap_samples"] = None
row["bootstrap_seed"] = None
row["bootstrap_block_width_trades"] = None
row["bootstrap_block_policy"] = None
row["bootstrap_limitation"] = None
if row["split"] == "test":
evaluated = predictions.filter(
(pl.col("split") == "test")
& (pl.col("model") == str(row["model"]))
& (pl.col("requested_model") == requested_model)
)
interval_seed = seed + index
interval = block_bootstrap_metric(
evaluated,
metric=metric,
block_column="bootstrap_block",
n_bootstrap=n_bootstrap,
seed=interval_seed,
)
row[f"{metric}_ci_low"] = interval.lower
row[f"{metric}_ci_high"] = interval.upper
row["bootstrap_status"] = interval.status
row["bootstrap_blocks"] = interval.n_blocks
row["bootstrap_samples"] = interval.n_bootstrap
row["bootstrap_seed"] = interval_seed
row["bootstrap_block_width_trades"] = block_width
row["bootstrap_block_policy"] = "fixed_contiguous_2x_label_horizon"
row["bootstrap_limitation"] = (
"percentile dependence diagnostic on fixed trade blocks; not a p-value or "
"confirmatory coverage guarantee"
)
rows.append(row)
return pl.DataFrame(rows, infer_schema_length=None).with_columns(
pl.lit(symbol).alias("symbol"),
pl.lit(symbol).alias("instrument"),
pl.lit(symbol).alias("instrument_scope"),
)
def _metric_delta_direction(metric: str) -> tuple[str, int]:
if metric in {"log_loss", "brier_score", "expected_calibration_error"}:
return "negative_selected_minus_prior_is_favorable", -1
if metric in {"accuracy", "balanced_accuracy", "roc_auc", "pr_auc"}:
return "positive_selected_minus_prior_is_favorable", 1
raise PublicPipelineError(f"unsupported paired hypothesis metric: {metric}")
def _paired_hypothesis_evaluation(
predictions: pl.DataFrame,
*,
symbol: str,
selected_model: str,
metric: str,
n_bootstrap: int,
seed: int,
horizon: int,
) -> dict[str, Any]:
selected = predictions.filter(
(pl.col("split") == "test") & (pl.col("requested_model") == selected_model)
)
baseline = predictions.filter(
(pl.col("split") == "test") & (pl.col("requested_model") == HYPOTHESIS_BASELINE)
)
if selected.is_empty():
raise PublicPipelineError(
f"validation-selected test predictions are missing for paired test: {symbol}"
)
if baseline.is_empty():
raise PublicPipelineError(f"historical-prior test predictions are missing for {symbol}")
identity_columns = [
"row_id",
"y_true",
"bootstrap_block",
"bootstrap_block_start_trade_id",
"bootstrap_block_end_trade_id",
"bootstrap_block_width_trades",
"bootstrap_block_policy",
]
for name, frame in (("selected", selected), ("historical_prior", baseline)):
missing = sorted(set(identity_columns).difference(frame.columns))
if missing:
raise PublicPipelineError(
f"{name} predictions lack paired-bootstrap identity columns: {missing}"
)
if frame.get_column("row_id").n_unique() != frame.height:
raise PublicPipelineError(f"{name} predictions contain duplicate row IDs for {symbol}")
selected_identity = selected.select(identity_columns).sort("row_id")
baseline_identity = baseline.select(identity_columns).sort("row_id")
if not selected_identity.equals(baseline_identity):
raise PublicPipelineError(
f"selected and historical-prior predictions do not share identical rows/blocks for {symbol}"
)
block_width = 2 * horizon
policies = selected.get_column("bootstrap_block_policy").unique().to_list()
widths = selected.get_column("bootstrap_block_width_trades").unique().to_list()
if policies != [HYPOTHESIS_BLOCK_POLICY] or widths != [block_width]:
raise PublicPipelineError(
f"paired hypothesis blocks do not implement the frozen 2x-horizon policy for {symbol}"
)
interval = paired_block_bootstrap_difference(
selected,
baseline,
metric=metric,
block_column="bootstrap_block",
n_bootstrap=n_bootstrap,
seed=seed,
)
favorable_direction, favorable_sign = _metric_delta_direction(metric)
signed_point = favorable_sign * interval.point_estimate
if not math.isfinite(signed_point):
point_assessment = "unavailable"
point_favorable: bool | None = None
elif signed_point > 0:
point_assessment = "favorable_point_only"
point_favorable = True
elif signed_point < 0:
point_assessment = "unfavorable_point"
point_favorable = False
else:
point_assessment = "point_tie"
point_favorable = False
if interval.lower is None or interval.upper is None:
interval_relation = "unavailable"
elif interval.lower <= 0.0 <= interval.upper:
interval_relation = "includes_zero"
elif favorable_sign * interval.lower > 0.0 and favorable_sign * interval.upper > 0.0:
interval_relation = "entirely_favorable"
else:
interval_relation = "entirely_unfavorable"
return {
"symbol": symbol,
"selected_model": selected_model,
"baseline": HYPOTHESIS_BASELINE,
"metric": metric,
"delta_definition": "selected_model_minus_historical_prior",
"point_delta": interval.point_estimate,
"ci_level": 0.95,
"ci_low": interval.lower,
"ci_high": interval.upper,
"n_obs": selected.height,
"n_blocks": interval.n_blocks,
"samples": interval.n_bootstrap,
"seed": interval.seed,
"status": interval.status,
"block_column": "bootstrap_block",
"block_policy": HYPOTHESIS_BLOCK_POLICY,
"block_width_trades": block_width,
"paired_row_ids_identical": True,
"paired_blocks_identical": True,
"favorable_direction": favorable_direction,
"point_favorable": point_favorable,
"point_assessment": point_assessment,
"interval_relation_to_zero": interval_relation,
"exploratory": True,
"significance_claim_authorized": False,
"h0_rejection_authorized": False,
"caveat": HYPOTHESIS_CAVEAT,
"cross_instrument_conclusion": CROSS_INSTRUMENT_CONCLUSION,
}
def _attach_paired_hypothesis_to_comparison(
comparison: pl.DataFrame,
hypothesis: Mapping[str, Any],
) -> pl.DataFrame:
selected_model = str(hypothesis["selected_model"])
rows: list[dict[str, Any]] = []
attached = 0
for source in comparison.to_dicts():
row = dict(source)
selected_test = row["split"] == "test" and row.get("requested_model") == selected_model
paired_fields = {
"paired_baseline": None,
"paired_metric": None,
"paired_metric_delta": None,
"paired_metric_delta_ci_low": None,
"paired_metric_delta_ci_high": None,
"paired_n_obs": None,
"paired_bootstrap_blocks": None,
"paired_bootstrap_samples": None,
"paired_bootstrap_seed": None,
"paired_bootstrap_status": None,
"paired_bootstrap_block_policy": None,
"paired_favorable_direction": None,
"paired_point_favorable": None,
"paired_exploratory": None,
"paired_significance_claim_authorized": None,
}
if selected_test:
attached += 1
paired_fields.update(
{
"paired_baseline": hypothesis["baseline"],
"paired_metric": hypothesis["metric"],
"paired_metric_delta": hypothesis["point_delta"],
"paired_metric_delta_ci_low": hypothesis["ci_low"],
"paired_metric_delta_ci_high": hypothesis["ci_high"],
"paired_n_obs": hypothesis["n_obs"],
"paired_bootstrap_blocks": hypothesis["n_blocks"],
"paired_bootstrap_samples": hypothesis["samples"],
"paired_bootstrap_seed": hypothesis["seed"],
"paired_bootstrap_status": hypothesis["status"],
"paired_bootstrap_block_policy": hypothesis["block_policy"],
"paired_favorable_direction": hypothesis["favorable_direction"],
"paired_point_favorable": hypothesis["point_favorable"],
"paired_exploratory": hypothesis["exploratory"],
"paired_significance_claim_authorized": hypothesis[
"significance_claim_authorized"
],
}
)
row.update(paired_fields)
rows.append(row)
if attached != 1:
raise PublicPipelineError(
"paired hypothesis metadata must attach to exactly one selected test comparison row"
)
return pl.DataFrame(rows, infer_schema_length=None)
def _rows_by_plan(
evaluation: pl.DataFrame, indices: np.ndarray[Any, np.dtype[np.int64]]
) -> pl.DataFrame:
indexed = evaluation.with_row_index("_research_row_id")
return indexed.filter(pl.col("_research_row_id").is_in(indices)).drop("_research_row_id")
def _evaluate_symbol(
trades: pl.DataFrame,
*,
config: ProjectConfig,
symbol_index: int,
) -> _SymbolResult:
symbol = str(trades.get_column("symbol")[0])
research = build_trade_only_research_frame(trades, config.features).with_columns(
pl.col("label_horizon_trades").alias("label_horizon_events")
)
temporal = validate_trade_only_temporal_contract(research)
evaluation = research.filter(pl.col("feature_ready"))
if evaluation.is_empty():
raise PublicPipelineError(f"trade-only feature construction produced no rows for {symbol}")
features = _trade_feature_columns(config, evaluation)
plan = expanding_walk_forward_splits(evaluation, config.evaluation)
ladder = evaluate_model_ladder(
evaluation,
plan,
config.models,
seed=config.run.seed + symbol_index * 100_000,
calibration_bins=config.evaluation.calibration_bins,
target="future_trade_up",
features=features,
)
block_width = 2 * config.features.label_horizon_events
predictions = _add_fixed_blocks(ladder.predictions, block_width=block_width)
comparison = _bootstrap_comparison(
ladder.comparison,
predictions,
symbol=symbol,
selected_model=ladder.selected_model,
metric=ladder.selection_metric,
n_bootstrap=config.evaluation.bootstrap_samples,
seed=config.run.seed + symbol_index * 100_000 + 10_000,
horizon=config.features.label_horizon_events,
)
selected = predictions.filter(
(pl.col("split") == "test") & (pl.col("requested_model") == ladder.selected_model)
)
if selected.is_empty() or selected.get_column("requested_model").n_unique() != 1:
raise PublicPipelineError(f"validation-selected test predictions are missing for {symbol}")
hypothesis = _paired_hypothesis_evaluation(
predictions,
symbol=symbol,
selected_model=ladder.selected_model,
metric=ladder.selection_metric,
n_bootstrap=config.evaluation.bootstrap_samples,
seed=config.run.seed + symbol_index * 100_000 + 20_000,
horizon=config.features.label_horizon_events,
)
comparison = _attach_paired_hypothesis_to_comparison(comparison, hypothesis)
return _SymbolResult(
symbol=symbol,
research=research,
evaluation=evaluation,
plan=plan,
predictions=predictions,
comparison=comparison,
selected_predictions=selected,
selected_model=ladder.selected_model,
feature_columns=features,
temporal_audit=asdict(temporal),
hypothesis_evaluation=hypothesis,
)
def _trade_summary(
trades: pl.DataFrame, continuity_audits: Sequence[Mapping[str, Any]]
) -> pl.DataFrame:
continuity = {
str(row["symbol"]): str(row["derived_continuity_id"]) for row in continuity_audits
}
rows: list[dict[str, Any]] = []
for frame in trades.partition_by("symbol", maintain_order=True):
symbol = str(frame.get_column("symbol")[0])
buy = frame.filter(pl.col("aggressor_side") == "buy")
sell = frame.filter(pl.col("aggressor_side") == "sell")
rows.append(
{
"symbol": symbol,
"rows": frame.height,
"first_trade_id": int(cast(int, frame.get_column("trade_id").min())),
"last_trade_id": int(cast(int, frame.get_column("trade_id").max())),
"observed_start_utc": _utc_from_ns(
int(cast(int, frame.get_column("event_ts_ns").min()))
),
"observed_end_utc": _utc_from_ns(
int(cast(int, frame.get_column("event_ts_ns").max()))
),
"total_quantity": float(cast(float, frame.get_column("quantity").sum())),
"total_quote_quantity": float(
cast(float, frame.get_column("quote_quantity").sum())
),
"buy_rows": buy.height,
"sell_rows": sell.height,
"buy_quantity": float(cast(float, buy.get_column("quantity").sum())),
"sell_quantity": float(cast(float, sell.get_column("quantity").sum())),
"derived_continuity_id": continuity[symbol],
"availability_basis": "exchange_event_time_proxy",
"analysis_kind": "public_trade_sample_summary_descriptive",
"descriptive_only": True,
}
)
return pl.DataFrame(rows).sort("symbol")
def _write_analyses(
*,
trades: pl.DataFrame,
symbol_results: Sequence[_SymbolResult],
continuity_audits: Sequence[Mapping[str, Any]],
config: ProjectConfig,
stage: Path,
generated_at_utc: str,
) -> dict[str, str]:
summary = _trade_summary(trades, continuity_audits)
stability_frames: list[pl.DataFrame] = []
flow_frames: list[pl.DataFrame] = []
flow_feature = f"trade_imbalance_w{max(config.features.trade_windows)}"
for result in symbol_results:
train = _rows_by_plan(result.evaluation, result.plan.final_train_indices)
test = _rows_by_plan(result.evaluation, result.plan.test_indices)
stability_frames.append(
feature_stability_summary(
train,
test,
feature_columns=result.feature_columns,
group_columns=("symbol",),
).with_columns(
pl.lit("final_training_period").alias("reference_period"),
pl.lit("untouched_final_test").alias("comparison_period"),
pl.lit(PUBLIC_EVIDENCE_TIER).alias("evidence_tier"),
)
)
labeled = result.evaluation.filter(~pl.col("right_censored"))
flow_frames.append(
ofi_future_return_association(
labeled,
horizon_return_columns={
config.features.label_horizon_events: "future_trade_return"
},
ofi_column=flow_feature,
min_observations=3,
).with_columns(
pl.lit("all_feature_ready_labeled_rows").alias("analysis_scope"),
pl.lit(PUBLIC_EVIDENCE_TIER).alias("evidence_tier"),
)
)
stability = pl.concat(stability_frames)
flow_return = pl.concat(flow_frames)
artifacts = {
"trade_summary": "analysis/trade_summary.parquet",
"feature_stability": "analysis/feature_stability.parquet",
"flow_return_analysis": "analysis/flow_return_analysis.parquet",
}
for name, frame in (
("trade_summary", summary),
("feature_stability", stability),
("flow_return_analysis", flow_return),
):
path = stage / artifacts[name]
path.parent.mkdir(parents=True, exist_ok=True)
frame.write_parquet(path)
_write_json(
stage / "analysis" / "manifest.json",
{
"generated_at_utc": generated_at_utc,
"evidence_tier": PUBLIC_EVIDENCE_TIER,
"descriptive_only": True,
"economic_claim_authorized": False,
"execution_claim_authorized": False,
"artifacts": {
"trade_summary": {"path": artifacts["trade_summary"], "rows": summary.height},
"feature_stability": {
"path": artifacts["feature_stability"],
"rows": stability.height,
"reference": "final training period",
"comparison": "untouched final test",
},
"flow_return_analysis": {
"path": artifacts["flow_return_analysis"],
"rows": flow_return.height,
"scope": "all feature-ready labeled rows",
},
},
"limitations": [
"Retrospective bounded sample; no confirmatory significance claim.",
"Exchange event time is an availability proxy, not local receipt evidence.",
"Trade-only observations cannot support book or execution analysis.",
],
},
)
return {**artifacts, "analysis_manifest": "analysis/manifest.json"}
def _quality_payload(
public: PublicTrades,
continuity_audits: Sequence[Mapping[str, Any]],
*,
generated_at_utc: str,
) -> dict[str, Any]:
return {
"generated_at_utc": generated_at_utc,
"dataset": "manifested_public_aggregate_trades",
"rows_checked": public.validation.rows_checked,
"summary": {
"errors": public.validation.error_count,
"warnings": public.validation.warning_count,
},
"normalized_schema_validation": {
"dataset": public.validation.dataset,
"rows_checked": public.validation.rows_checked,
"errors": public.validation.error_count,
"warnings": public.validation.warning_count,
"findings": [asdict(finding) for finding in public.validation.findings],
},
"aggregate_trade_continuity": list(continuity_audits),
"row_bound": public.row_bound,
"row_bound_respected": public.rows <= public.row_bound,
"canonical_source_order": list(public.canonical_order),
"mutation_policy": (
"verified normalized observations were not repaired or overwritten; continuity "
"exists only in persisted derived research frames"
),
}
def _hypothesis_number(value: object) -> str:
if value is None:
return "N/A"
if isinstance(value, (int, float)) and not isinstance(value, bool):
number = float(value)
return f"{number:.6f}" if math.isfinite(number) else "N/A"
return str(value)
def _render_hypothesis_report_section(stage: Path) -> str:
artifact = stage / "metrics" / "hypothesis_evaluation.json"
payload = _mapping(read_json(artifact))
if payload is None:
raise PublicPipelineError("serialized hypothesis evaluation must be a JSON object")
rows = payload.get("per_symbol")
if not isinstance(rows, list) or not rows:
raise PublicPipelineError("serialized hypothesis evaluation has no per-symbol rows")
lines = [
"## Paired H0/H1 diagnostic",
"",
(
"The frozen comparison is validation-selected test predictions versus the "
"historical-prior baseline on identical `row_id` and fixed 2x-horizon blocks. "
"For Δ log-loss (selected minus historical prior), a negative value favors the "
"selected model."
),
"",
(
"| Symbol | Selected model | Baseline | Metric | Point Δ | Paired 95% interval | "
"Observations | Blocks | Samples | Seed | Status |"
),
"| --- | --- | --- | --- | ---: | --- | ---: | ---: | ---: | ---: | --- |",
]
for value in rows:
row = _mapping(value)
if row is None:
raise PublicPipelineError("serialized per-symbol hypothesis row is not an object")
interval = (
f"[{_hypothesis_number(row.get('ci_low'))}, {_hypothesis_number(row.get('ci_high'))}]"
)
cells = (
row.get("symbol"),
row.get("selected_model"),
row.get("baseline"),
row.get("metric"),
_hypothesis_number(row.get("point_delta")),
interval,
row.get("n_obs"),
row.get("n_blocks"),
row.get("samples"),
row.get("seed"),
row.get("status"),
)
lines.append(
"| "
+ " | ".join(str(cell).replace("|", "\\|").replace("\n", " ") for cell in cells)
+ " |"
)
lines.extend(
[
"",
(
"These paired percentile intervals are dependence diagnostics, not p-values "
"or confirmatory significance intervals; H0 is not rejected by this "
"exploratory design."
),
"",
(
"No cross-instrument estimate was pooled. Mixed directions, if present, "
"cannot support persistent alpha; even matching directions would remain "
"bounded, sample-specific diagnostics."
),
]
)
return "\n".join(lines)
def _report_set(stage: Path) -> None:
bundle = load_run_bundle(stage, require_complete=False, verify_integrity=False)
hypothesis_section = _render_hypothesis_report_section(stage)
_atomic_write_text(
stage / "reports" / "technical_report.md",
render_technical_report(bundle),
)
_atomic_write_text(
stage / "reports" / "executive_memo.md",
render_executive_memo(bundle),
)
_atomic_write_text(
stage / "reports" / "model_comparison.md",
render_model_comparison(bundle)
+ "\nExecution status: `NOT_RUN`. Reason: "
+ EXECUTION_EXCLUSION_REASON
+ "\n\n"
+ hypothesis_section
+ "\n",
)
def produce_public_trade_run(
config: ProjectConfig,
stage: Path,
*,
ingestion_manifest_path: str | Path,
ingestion_manifest_sha256: str,
) -> None:
"""Populate an empty atomic stage with one verified public trade-only run.
The caller owns staging-directory creation, cleanup on failure, and final
rename. This function writes ``_SUCCESS`` only after every artifact, report,
and checksum is durable in the supplied stage.
"""
destination = stage.resolve()
if destination.exists() and any(destination.iterdir()):
raise PublicPipelineError("public run stage must be empty")
destination.mkdir(parents=True, exist_ok=True)
if config.data.mode != "binance_rest":
raise PublicPipelineError("public trade producer requires data.mode='binance_rest'")
explicit_manifest_path = Path(ingestion_manifest_path).resolve()
public = read_public_trades(
config,
explicit_manifest_path,
ingestion_manifest_sha256=ingestion_manifest_sha256,
)
if public.evidence_tier == "FULL_DATA" and not public.all_requested_ranges_complete:
raise PublicPipelineError("reader evidence tier contradicts manifested completeness")
derived_trades, continuity_audits = _derive_verified_continuity(public, config)
if public.validation.has_errors:
raise PublicPipelineError("verified public normalized input has quality errors")
generated = provenance_header(
project_root=config.project_root,
config_hash=config.hash,
evidence_tier=PUBLIC_EVIDENCE_TIER,
input_manifests=[],
)
generated_at = cast(str, generated["generated_at_utc"])
lineage = _input_lineage(public, config)
manifest_hashes = cast(list[str], lineage["manifest_sha256"])
data_hashes = cast(list[str], lineage["data_sha256"])
protocol_path, protocol_sha256 = _freeze_protocol(config, destination)
symbol_results: list[_SymbolResult] = []
for symbol_index, symbol in enumerate(config.data.symbols):
symbol_results.append(
_evaluate_symbol(
derived_trades.filter(pl.col("symbol") == symbol),
config=config,
symbol_index=symbol_index,
)
)
research_root = destination / "research"
model_root = destination / "models"
combined_research: list[pl.DataFrame] = []
combined_evaluation: list[pl.DataFrame] = []
combined_predictions: list[pl.DataFrame] = []
combined_comparison: list[pl.DataFrame] = []
combined_selected: list[pl.DataFrame] = []
symbol_manifest: dict[str, Any] = {}
for result in symbol_results:
slug = result.symbol.lower()
symbol_research = research_root / slug
symbol_models = model_root / slug
symbol_research.mkdir(parents=True, exist_ok=True)
symbol_models.mkdir(parents=True, exist_ok=True)
result.research.write_parquet(symbol_research / "research_frame.parquet")
result.evaluation.write_parquet(symbol_research / "evaluation_frame.parquet")
_write_json(symbol_research / "folds.json", _serialize_plan(result.plan))
result.predictions.write_parquet(symbol_models / "predictions.parquet")
result.comparison.write_parquet(symbol_models / "comparison.parquet")
result.selected_predictions.write_parquet(
symbol_models / "selected_test_predictions.parquet"
)
combined_research.append(result.research)
combined_evaluation.append(result.evaluation)
combined_predictions.append(result.predictions)
combined_comparison.append(result.comparison)
combined_selected.append(result.selected_predictions)
symbol_manifest[result.symbol] = {
"research_frame": f"research/{slug}/research_frame.parquet",
"evaluation_frame": f"research/{slug}/evaluation_frame.parquet",
"folds": f"research/{slug}/folds.json",
"predictions": f"models/{slug}/predictions.parquet",
"comparison": f"models/{slug}/comparison.parquet",
"selected_test_predictions": f"models/{slug}/selected_test_predictions.parquet",
"selected_model": result.selected_model,
"selection_metric": config.models.selection_metric,
"selection_source": "validation folds only",
"test_used_for_selection": False,
"feature_columns": list(result.feature_columns),
"temporal_audit": dict(result.temporal_audit),
"paired_hypothesis_evaluation": dict(result.hypothesis_evaluation),
"test_start_utc": _utc_from_ns(result.plan.test_start_ts_ns),
"test_end_utc": _utc_from_ns(result.plan.test_end_ts_ns),
}
research = pl.concat(combined_research)
evaluation = pl.concat(combined_evaluation)
predictions = pl.concat(combined_predictions)
comparison = pl.concat(combined_comparison)
selected_predictions = pl.concat(combined_selected)
research.write_parquet(research_root / "research_frame.parquet")
evaluation.write_parquet(research_root / "evaluation_frame.parquet")
predictions.write_parquet(model_root / "predictions.parquet")
comparison.write_parquet(model_root / "comparison.parquet")
selected_predictions.write_parquet(model_root / "selected_test_predictions.parquet")
metrics_root = destination / "metrics"
_write_json(metrics_root / "predictive_metrics.json", comparison.to_dicts())
hypothesis_rows = [dict(result.hypothesis_evaluation) for result in symbol_results]
hypothesis_payload = {
"schema_version": PUBLIC_PIPELINE_SCHEMA_VERSION,
"generated_at_utc": generated_at,
"evidence_tier": PUBLIC_EVIDENCE_TIER,
"hypotheses": {
"H0": (
"The validation-selected model does not improve held-out selection-metric "
"performance over the historical-prior classifier."
),
"H1_exploratory": (
"Causal aggregate-trade features improve held-out selection-metric "
"performance relative to the historical prior."
),
},
"comparison_contract": (
"per-symbol validation-selected final-test predictions minus historical-prior "
"predictions on identical row_id and fixed dependency block"
),
"selection_metric": config.models.selection_metric,
"delta_definition": "selected_model_minus_historical_prior",
"bootstrap": {
"method": "paired fixed-block percentile bootstrap",
"ci_level": 0.95,
"samples": config.evaluation.bootstrap_samples,
"seed_policy": "run_seed + symbol_index*100000 + 20000",
"block_policy": HYPOTHESIS_BLOCK_POLICY,
"block_width_trades": 2 * config.features.label_horizon_events,
},
"per_symbol": hypothesis_rows,
"cross_instrument_conclusion": {
"status": "not_inferred",
"pooling_performed": False,
"persistent_alpha_claim_authorized": False,
"text": CROSS_INSTRUMENT_CONCLUSION,
},
"exploratory": True,
"significance_claim_authorized": False,
"caveat": HYPOTHESIS_CAVEAT,
}
_write_json(metrics_root / "hypothesis_evaluation.json", hypothesis_payload)
_write_json(metrics_root / "execution_metrics.json", [])
_write_json(metrics_root / "execution_sensitivity.json", [])
_write_json(
metrics_root / "execution_exclusion.json",
{
"status": "NOT_RUN",
"reason": EXECUTION_EXCLUSION_REASON,
"execution_metrics_rows": 0,
"execution_sensitivity_rows": 0,
"pnl_calculated": False,
"profitability_claim_authorized": False,
},
)
analysis_artifacts = _write_analyses(
trades=derived_trades,
symbol_results=symbol_results,
continuity_audits=continuity_audits,
config=config,
stage=destination,
generated_at_utc=generated_at,
)
quality = _quality_payload(public, continuity_audits, generated_at_utc=generated_at)
_write_json(destination / "quality" / "summary.json", quality)
market_state = evaluation.select(
"symbol",
"decision_ts_ns",
"decision_trade_id",
"price",
"quantity",
"aggressor_side",
f"trade_imbalance_w{max(config.features.trade_windows)}",
f"realized_volatility_w{config.features.volatility_window}",
pl.lit(PUBLIC_EVIDENCE_TIER).alias("evidence_tier"),
pl.lit("trade_only_no_book_state").alias("market_state_scope"),
).sort(["decision_ts_ns", "symbol", "decision_trade_id"])
dashboard_path = destination / "dashboard" / "market_state.parquet"
dashboard_path.parent.mkdir(parents=True, exist_ok=True)
market_state.write_parquet(dashboard_path)
data_snapshot = {
"schema_version": PUBLIC_PIPELINE_SCHEMA_VERSION,
"mode": "binance_rest_trade_only",
"source": config.data.source,
"evidence_tier": PUBLIC_EVIDENCE_TIER,
"reader_effective_evidence_tier": public.evidence_tier,
"configured_requested_evidence_tier": config.run.evidence_tier,
"producer_evidence_policy": (
"always PUBLIC_SAMPLE_PARTIAL for this retrospective exploratory protocol"
),
"requested_period_utc": {
"start": config.data.start.isoformat().replace("+00:00", "Z"),
"end": config.data.end.isoformat().replace("+00:00", "Z") if config.data.end else None,
},
"observed_period_utc": {
"start": public.observed.start_utc,
"end_inclusive": public.observed.end_inclusive_utc,
},
"rows": public.rows,
"row_bound": public.row_bound,
"all_requested_ranges_complete": public.all_requested_ranges_complete,
"canonical_order": list(public.canonical_order),
"manifest_authority": {
"policy": "explicit path and caller-supplied SHA-256; no directory discovery",
"absolute_path": str(public.ingestion_manifest_path.resolve()),
"project_relative_or_absolute_path": _project_path(
public.ingestion_manifest_path, config.project_root
),
"sha256": public.ingestion_manifest_sha256,
},
"lineage": lineage,
"symbols": [
{
"symbol": item.symbol,
"rows": item.rows,
"complete_range": item.complete_range,
"tick_size": str(item.tick_size),
"lot_size": str(item.lot_size),
"observed_start_utc": item.observed.start_utc,
"observed_end_inclusive_utc": item.observed.end_inclusive_utc,
}
for item in public.symbols
],
"transformation": (
"source normalized rows unchanged; one derived continuity epoch per symbol was "
"assigned only after contiguous aggregate-ID and nonreversing-clock checks"
),
}
_write_json(destination / "data" / "manifest_snapshot.json", data_snapshot)
resolved_config = config.public_dict()
resolved_config["effective_evidence_tier"] = PUBLIC_EVIDENCE_TIER
resolved_config["reader_effective_evidence_tier"] = public.evidence_tier
resolved_config["evidence_policy"] = (
"retrospective public trade protocol cannot be promoted above PUBLIC_SAMPLE_PARTIAL"
)
resolved_config["protocol"] = {
"path": protocol_path,
"sha256": protocol_sha256,
}
_write_json(destination / "resolved_config.json", resolved_config)
git_metadata = cast(Mapping[str, Any], generated["git"])
run_key_inputs = {
"config_sha256": config.hash,
"input_manifest_sha256": manifest_hashes,
"input_data_sha256": data_hashes,
"protocol_sha256": protocol_sha256,
"git": {
"commit": str(git_metadata.get("commit", "UNKNOWN")),
"dirty": bool(git_metadata.get("dirty", False)),
"source_tree_sha256": str(git_metadata.get("source_tree_sha256", "UNKNOWN")),
},
"seed": config.run.seed,
"protocol": "public_aggregate_trade_exploratory_v1",
}
run_key = _stable_sha256(run_key_inputs)
generated.update(
{
"evidence_tier": PUBLIC_EVIDENCE_TIER,
"requested_evidence_tier": config.run.evidence_tier,
"effective_evidence_tier": PUBLIC_EVIDENCE_TIER,
"reader_effective_evidence_tier": public.evidence_tier,
"input_manifest_sha256": manifest_hashes,
"input_data_sha256": data_hashes,
"ingestion_manifest_path": _project_path(
public.ingestion_manifest_path, config.project_root
),
"ingestion_manifest_absolute_path": str(public.ingestion_manifest_path.resolve()),
"ingestion_manifest_sha256": public.ingestion_manifest_sha256,
"ingestion_manifest_authority": (
"explicit path and caller-supplied SHA-256; no directory discovery"
),
"protocol_path": protocol_path,
"protocol_sha256": protocol_sha256,
"run_key": run_key,
"run_key_inputs": run_key_inputs,
"pipeline_schema_version": PUBLIC_PIPELINE_SCHEMA_VERSION,
"seed": config.run.seed,
"observed_start_utc": public.observed.start_utc,
"observed_end_utc": public.observed.end_inclusive_utc,
"data_availability_clock": "exchange_event_time_proxy",
"local_receipt_time_available": False,
"execution_simulated": False,
}
)
_write_json(destination / "provenance.json", generated)
artifacts = {
"resolved_config": "resolved_config.json",
"protocol": protocol_path,
"data_manifest_snapshot": "data/manifest_snapshot.json",
"quality_summary": "quality/summary.json",
"research_frame": "research/research_frame.parquet",
"evaluation_frame": "research/evaluation_frame.parquet",
"predictions": "models/predictions.parquet",
"selected_test_predictions": "models/selected_test_predictions.parquet",
"model_comparison_data": "models/comparison.parquet",
"predictive_metrics": "metrics/predictive_metrics.json",
"hypothesis_evaluation": "metrics/hypothesis_evaluation.json",
"execution_metrics": "metrics/execution_metrics.json",
"execution_sensitivity": "metrics/execution_sensitivity.json",
"execution_exclusion": "metrics/execution_exclusion.json",
"market_state": "dashboard/market_state.parquet",
"technical_report": "reports/technical_report.md",
"executive_memo": "reports/executive_memo.md",
"model_comparison": "reports/model_comparison.md",
**analysis_artifacts,
}
run_manifest = {
"schema_version": PUBLIC_PIPELINE_SCHEMA_VERSION,
"run_id": config.run.name,
"run_key": run_key,
"status": "complete",
"evidence_tier": PUBLIC_EVIDENCE_TIER,
"data": {
"mode": "binance_rest_trade_only",
"source": config.data.source,
"symbols": list(config.data.symbols),
"rows": public.rows,
"row_bound": public.row_bound,
"all_requested_ranges_complete": public.all_requested_ranges_complete,
"reader_effective_evidence_tier": public.evidence_tier,
"configured_requested_evidence_tier": config.run.evidence_tier,
"observed_start_utc": public.observed.start_utc,
"observed_end_utc": public.observed.end_inclusive_utc,
"observed_start_ts_ns": public.observed.start_ns,
"observed_end_ts_ns": public.observed.end_inclusive_ns,
"availability_basis": "exchange_event_time_proxy",
"local_receipt_time_available": False,
"symbol_coverage": [
{
"symbol": item.symbol,
"rows": item.rows,
"complete_range": item.complete_range,
"observed_start_utc": item.observed.start_utc,
"observed_end_inclusive_utc": item.observed.end_inclusive_utc,
}
for item in public.symbols
],
},
"artifacts": artifacts,
"research": {
"question": (
"whether recent observable aggregate-trade direction and size contain "
"out-of-time information about future trade-price direction"
),
"target": "future_trade_up",
"label_horizon_trades": config.features.label_horizon_events,
"evaluation_contract": "separate per-symbol expanding purged walk-forward",
"selection_contract": "validation folds only; final test never used for selection",
"bootstrap_contract": {
"samples": config.evaluation.bootstrap_samples,
"seeded": True,
"block_width_trades": 2 * config.features.label_horizon_events,
"block_policy": "fixed_contiguous_2x_label_horizon",
"status": "dependence diagnostic, not confirmatory significance",
},
"hypothesis_evaluation": {
"artifact": "metrics/hypothesis_evaluation.json",
"hypotheses": ["H0", "H1_exploratory"],
"baseline": HYPOTHESIS_BASELINE,
"metric": config.models.selection_metric,
"delta_definition": "selected_model_minus_historical_prior",
"paired_on": ["row_id", "bootstrap_block"],
"per_symbol_only": True,
"cross_instrument_pooling": False,
"persistent_alpha_claim_authorized": False,
"exploratory": True,
"significance_claim_authorized": False,
"caveat": HYPOTHESIS_CAVEAT,
"cross_instrument_conclusion": CROSS_INSTRUMENT_CONCLUSION,
},
"symbols": symbol_manifest,
"descriptive_analysis": {
"manifest": analysis_artifacts["analysis_manifest"],
"descriptive_only": True,
"economic_claim_authorized": False,
},
},
"execution_assumptions": {
"status": "NOT_RUN",
"reason": EXECUTION_EXCLUSION_REASON,
"pnl_calculated": False,
"fills_calculated": False,
"capacity_calculated": False,
"profitability_claim_authorized": False,
},
"warnings": [
"PUBLIC_SAMPLE_PARTIAL: results are bounded, retrospective, and sample-specific.",
"Exchange event time is an availability proxy and not local receipt evidence.",
EXECUTION_EXCLUSION_REASON,
"Bootstrap intervals are dependence diagnostics and not significance claims.",
CROSS_INSTRUMENT_CONCLUSION,
],
}
_write_json(destination / "run_manifest.json", run_manifest)
_report_set(destination)
if _input_lineage(public, config) != lineage:
raise PublicPipelineError("external input lineage changed while producing the run")
write_checksum_manifest(destination)
descriptor = os.open(destination / "_SUCCESS", os.O_CREAT | os.O_EXCL | os.O_WRONLY)
with os.fdopen(descriptor, "w", encoding="utf-8") as success:
success.write("complete\n")
load_run_bundle(destination)
|