Datasets:
Tasks:
Tabular Classification
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
economics
quantitative-finance
causal-inference
macroeconomics
housing-economics
market-microstructure
License:
File size: 35,503 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 | """One explicit four-day public aggregate-trade exploratory study.
This producer is deliberately separate from the frozen M8 trade and live-L2
authorities. It reuses their bounded archive, normalization, causal-feature,
and numeric fitted-state primitives without weakening either frozen contract.
"""
from __future__ import annotations
import argparse
import hashlib
import json
import os
import shutil
import tempfile
import tomllib
from collections.abc import Mapping, Sequence
from dataclasses import asdict
from datetime import date
from pathlib import Path
from typing import Any, cast
import polars as pl
from microstructure.config import FeatureConfig, ModelConfig
from microstructure.data.binance import BinancePublicClient, SymbolMetadata
from microstructure.data.binance_archive import (
AcquiredDailyArchive,
ArchiveDownloadLimits,
BinanceArchiveClient,
DailyArchiveRequest,
)
from microstructure.data.evidence_budget import RetainedEvidenceBudget
from microstructure.m8_config import (
M8Claims,
M8Features,
M8Models,
M8Period,
M8Quality,
M8Study,
M8StudyConfig,
)
from microstructure.m8_manifest import M8ArchiveEntry, M8SymbolMetadata
from microstructure.m8_normalization import normalize_m8_archive
from microstructure.provenance import (
git_source_tree_sha256,
read_json,
sha256_file,
strict_git_state,
utc_now_iso,
)
from microstructure.research.multidate import (
AnalysisLock,
LockedSelection,
evaluate_locked_multidate_tests,
select_multidate_model,
)
from microstructure.research.trade_only import (
build_trade_only_research_frame,
validate_trade_only_temporal_contract,
)
SCHEMA_VERSION = "exploratory-aggtrades-study-v1"
EVIDENCE_TIER = "PUBLIC_ARCHIVE_EXPLORATORY"
EXPECTED_DATES = (
("2026-08-05", "train"),
("2026-08-06", "validation"),
("2026-08-07", "primary_test"),
("2026-08-08", "replication_test"),
)
EXPECTED_SYMBOLS = ("BTCUSDT", "ETHUSDT")
PROTOCOL_RELATIVE_PATH = "docs/EXPLORATORY_AGGTRADES_2026_08_05_08.md"
SUCCESS_BYTES = b"complete\n"
class ExploratoryStudyError(RuntimeError):
"""Raised when the exploratory producer cannot preserve its authority."""
def _canonical_bytes(value: object) -> bytes:
return json.dumps(
value,
sort_keys=True,
separators=(",", ":"),
allow_nan=False,
).encode("utf-8")
def _sha_bytes(value: bytes) -> str:
return hashlib.sha256(value).hexdigest()
def _atomic_write(path: Path, content: bytes) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
descriptor, raw_temporary = tempfile.mkstemp(prefix=f".{path.name}.", dir=path.parent)
temporary = Path(raw_temporary)
try:
with os.fdopen(descriptor, "wb") as handle:
handle.write(content)
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary, path)
parent = os.open(path.parent, os.O_RDONLY)
try:
os.fsync(parent)
finally:
os.close(parent)
finally:
temporary.unlink(missing_ok=True)
def _write_json(path: Path, value: object) -> None:
_atomic_write(path, _canonical_bytes(value) + b"\n")
def _mapping(value: object, label: str) -> Mapping[str, Any]:
if not isinstance(value, Mapping) or not all(isinstance(key, str) for key in value):
raise ExploratoryStudyError(f"{label} must be a string-keyed table")
return cast(Mapping[str, Any], value)
def _exact_keys(value: Mapping[str, Any], expected: set[str], label: str) -> None:
if set(value) != expected:
raise ExploratoryStudyError(f"{label} keys differ from the exploratory contract")
def _load_config(path: Path) -> M8StudyConfig:
source_path = path.resolve()
source = source_path.read_bytes()
try:
payload = tomllib.loads(source.decode("utf-8"))
except (UnicodeDecodeError, tomllib.TOMLDecodeError) as error:
raise ExploratoryStudyError("cannot parse exploratory configuration") from error
root = _mapping(payload, "configuration")
_exact_keys(root, {"study", "periods", "features", "models", "quality", "claims"}, "config")
study_raw = _mapping(root["study"], "study")
periods_raw = root["periods"]
features_raw = _mapping(root["features"], "features")
models_raw = _mapping(root["models"], "models")
quality_raw = _mapping(root["quality"], "quality")
claims_raw = _mapping(root["claims"], "claims")
if not isinstance(periods_raw, list) or len(periods_raw) != 4:
raise ExploratoryStudyError("periods must declare exactly four dates")
study = M8Study(
name=str(study_raw["name"]),
protocol_version=str(study_raw["protocol_version"]),
evidence_tier=str(study_raw["evidence_tier"]),
seed=int(study_raw["seed"]),
source=str(study_raw["source"]),
symbols=tuple(str(value) for value in cast(Sequence[object], study_raw["symbols"])),
selection_metric=str(study_raw["selection_metric"]),
target=str(study_raw["target"]),
label_horizon_events=int(study_raw["label_horizon_events"]),
calibration_fraction=float(study_raw["calibration_fraction"]),
bootstrap_samples=int(study_raw["bootstrap_samples"]),
bootstrap_block_events=int(study_raw["bootstrap_block_events"]),
feature_stability_bins=int(study_raw["feature_stability_bins"]),
max_archive_compressed_bytes=int(study_raw["max_archive_compressed_bytes"]),
max_archive_uncompressed_bytes=int(study_raw["max_archive_uncompressed_bytes"]),
max_total_download_bytes=int(study_raw["max_total_download_bytes"]),
)
periods = tuple(
M8Period(
date=date.fromisoformat(str(_mapping(value, "period")["date"])),
role=cast(Any, str(_mapping(value, "period")["role"])),
)
for value in periods_raw
)
features = M8Features(
trade_windows=tuple(
int(cast(Any, value)) for value in cast(Sequence[object], features_raw["trade_windows"])
),
volatility_window=int(features_raw["volatility_window"]),
intensity_window=int(features_raw["intensity_window"]),
large_trade_quantile=float(features_raw["large_trade_quantile"]),
)
models = M8Models(
logistic_c_values=tuple(
float(cast(Any, value))
for value in cast(Sequence[object], models_raw["logistic_c_values"])
),
tree_max_depth_values=tuple(
int(cast(Any, value))
for value in cast(Sequence[object], models_raw["tree_max_depth_values"])
),
tree_min_samples_leaf=int(models_raw["tree_min_samples_leaf"]),
)
quality = M8Quality(
fail_on_error=bool(quality_raw["fail_on_error"]),
require_complete_daily_archive=bool(quality_raw["require_complete_daily_archive"]),
require_contiguous_trade_ids_within_symbol_date=bool(
quality_raw["require_contiguous_trade_ids_within_symbol_date"]
),
require_nondecreasing_event_time=bool(quality_raw["require_nondecreasing_event_time"]),
allow_quality_warnings=bool(quality_raw["allow_quality_warnings"]),
)
claims = M8Claims(
allow_p_values=bool(claims_raw["allow_p_values"]),
allow_significance_claim=bool(claims_raw["allow_significance_claim"]),
allow_cross_instrument_pooling=bool(claims_raw["allow_cross_instrument_pooling"]),
allow_execution_claim=bool(claims_raw["allow_execution_claim"]),
allow_profitability_claim=bool(claims_raw["allow_profitability_claim"]),
)
config = M8StudyConfig(
path=source_path,
source_sha256=_sha_bytes(source),
study=study,
periods=periods,
features=features,
models=models,
quality=quality,
claims=claims,
)
observed_dates = tuple((item.date.isoformat(), item.role) for item in config.periods)
if (
config.study.name != "binance-aggtrades-2026-08-05-08-exploratory"
or config.study.protocol_version != "1.0.0"
or config.study.evidence_tier != EVIDENCE_TIER
or config.study.source != "binance_spot_daily_aggtrades_archive"
or config.study.symbols != EXPECTED_SYMBOLS
or observed_dates != EXPECTED_DATES
or config.study.selection_metric != "log_loss"
or config.study.target != "future_trade_up"
or config.study.label_horizon_events != 20
or not config.quality.fail_on_error
or not config.quality.allow_quality_warnings
or any(asdict(config.claims).values())
):
raise ExploratoryStudyError("configuration differs from the declared exploratory study")
if config.study.max_total_download_bytes < 1:
raise ExploratoryStudyError("total retained-evidence budget must be positive")
return config
def _protocol_path(config: M8StudyConfig) -> Path:
result = config.path.parent.parent / PROTOCOL_RELATIVE_PATH
if not result.is_file() or result.is_symlink():
raise ExploratoryStudyError("exploratory protocol document is missing or symbolic")
return result
def _metadata_authority(value: SymbolMetadata) -> M8SymbolMetadata:
sidecar = _mapping(read_json(value.source_manifest_path), "exchangeInfo sidecar")
source_uri = sidecar.get("source_uri")
if type(source_uri) is not str or not source_uri:
raise ExploratoryStudyError("exchangeInfo sidecar lacks its source URI")
return M8SymbolMetadata(
symbol=value.symbol,
status=value.status,
tick_size=value.tick_size,
lot_size=value.lot_size,
observed_ts_ns=value.observed_ts_ns,
raw_path=value.source_path.resolve(),
raw_sha256=sha256_file(value.source_path),
raw_bytes=value.source_path.stat().st_size,
source_uri=source_uri,
source_manifest_path=value.source_manifest_path.resolve(),
source_manifest_sha256=sha256_file(value.source_manifest_path),
source_manifest_bytes=value.source_manifest_path.stat().st_size,
)
def _feature_config(config: M8StudyConfig) -> FeatureConfig:
return FeatureConfig(
trade_windows=config.features.trade_windows,
volatility_window=config.features.volatility_window,
intensity_window=config.features.intensity_window,
label_horizon_events=config.study.label_horizon_events,
large_trade_quantile=config.features.large_trade_quantile,
)
def _model_config(config: M8StudyConfig) -> ModelConfig:
return ModelConfig(
selection_metric=config.study.selection_metric,
logistic_c_values=config.models.logistic_c_values,
tree_max_depth_values=config.models.tree_max_depth_values,
tree_min_samples_leaf=config.models.tree_min_samples_leaf,
)
def _feature_columns(config: M8StudyConfig) -> tuple[str, ...]:
columns = ["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}",
)
)
columns.extend(
(
f"trade_count_w{config.features.intensity_window}",
f"trade_intensity_w{config.features.intensity_window}",
f"realized_volatility_w{config.features.volatility_window}",
)
)
return tuple(dict.fromkeys(columns))
def _evaluation_columns(config: M8StudyConfig) -> tuple[str, ...]:
return (
"study_date",
"study_role",
"symbol",
"decision_ts_ns",
"decision_sequence",
"decision_trade_id",
"continuity_id",
"feature_continuity_id",
"label_continuity_id",
"max_feature_source_ts_ns",
"max_feature_source_trade_id",
"label_start_ts_ns",
"label_start_trade_id",
"label_information_end_ts_ns",
"label_information_end_trade_id",
"feature_ready",
"right_censored",
config.study.target,
*_feature_columns(config),
)
def _entry_payload(entry: M8ArchiveEntry, root: Path) -> dict[str, object]:
return {
"symbol": entry.symbol,
"date": entry.date.isoformat(),
"role": entry.role,
"rows": entry.rows,
"first_trade_id": entry.first_trade_id,
"last_trade_id": entry.last_trade_id,
"observed_start_ns": entry.observed_start_ns,
"observed_end_inclusive_ns": entry.observed_end_inclusive_ns,
"quality_errors": entry.quality_errors,
"quality_warnings": entry.quality_warnings,
"raw_zip": {
"path": str(entry.raw_zip_path.relative_to(root)),
"sha256": entry.raw_zip_sha256,
"bytes": entry.raw_zip_bytes,
},
"official_checksum": {
"path": str(entry.raw_checksum_path.relative_to(root)),
"sha256": entry.raw_checksum_sha256,
"bytes": entry.raw_checksum_bytes,
},
"normalized_manifest": {
"path": str(entry.normalized_dataset_manifest_path.relative_to(root)),
"sha256": entry.normalized_dataset_manifest_sha256,
"bytes": entry.normalized_dataset_manifest_bytes,
},
"parts": [
{
"path": str(item.data_path.relative_to(root)),
"sha256": item.data_sha256,
"bytes": item.data_bytes,
"rows": item.rows,
}
for item in entry.normalized_parts
],
"quality_report": {
"path": str(entry.quality_report_path.relative_to(root)),
"sha256": entry.quality_report_sha256,
"bytes": entry.quality_report_bytes,
},
"quality_findings": {
"path": str(entry.quality_findings_path.relative_to(root)),
"sha256": entry.quality_findings_sha256,
"bytes": entry.quality_findings_bytes,
},
}
def _build_evaluation_frame(
entry: M8ArchiveEntry,
config: M8StudyConfig,
data_root: Path,
) -> Path:
paths = [item.data_path for item in entry.normalized_parts]
frame = pl.read_parquet(paths, rechunk=False)
research = build_trade_only_research_frame(frame, _feature_config(config)).with_columns(
pl.lit(entry.date.isoformat()).alias("study_date"),
pl.lit(entry.role).alias("study_role"),
)
validate_trade_only_temporal_contract(research)
result = research.select(_evaluation_columns(config))
output = data_root / "derived" / "research" / entry.symbol / entry.date.isoformat()
output.mkdir(parents=True, exist_ok=True)
destination = output / "evaluation.parquet"
result.write_parquet(destination, compression="zstd", statistics=True)
del frame, research, result
return destination
def _source_authority(project_root: Path) -> dict[str, object]:
state = strict_git_state(project_root)
if state.dirty:
raise ExploratoryStudyError("exploratory producer requires a clean committed source")
return {
"commit": state.commit,
"dirty": False,
"source_tree_sha256": git_source_tree_sha256(project_root),
}
def _raw_manifest(
config: M8StudyConfig,
metadata: Mapping[str, M8SymbolMetadata],
archives: Mapping[tuple[str, str], AcquiredDailyArchive],
data_root: Path,
) -> tuple[Path, str]:
raw_root = data_root / "raw"
payload = {
"schema_version": "exploratory-aggtrades-raw-v1",
"config_sha256": config.hash,
"config_source_sha256": config.source_sha256,
"protocol_sha256": sha256_file(_protocol_path(config)),
"csv_members_opened": False,
"metadata": [
{
"symbol": symbol,
"status": item.status,
"tick_size": str(item.tick_size),
"lot_size": str(item.lot_size),
"path": str(item.raw_path.relative_to(raw_root)),
"sha256": item.raw_sha256,
"bytes": item.raw_bytes,
"sidecar_path": str(item.source_manifest_path.relative_to(raw_root)),
"sidecar_sha256": item.source_manifest_sha256,
"sidecar_bytes": item.source_manifest_bytes,
}
for symbol, item in sorted(metadata.items())
],
"archives": [
{
"symbol": symbol,
"date": day,
"role": next(item.role for item in config.periods if item.date.isoformat() == day),
"zip_path": str(value.archive_artifact.path.relative_to(raw_root)),
"zip_sha256": value.archive_artifact.sha256,
"zip_bytes": value.archive_artifact.bytes,
"zip_sidecar_path": str(value.archive_artifact.manifest_path.relative_to(raw_root)),
"zip_sidecar_sha256": value.archive_artifact.manifest_sha256,
"checksum_path": str(value.checksum_artifact.path.relative_to(raw_root)),
"checksum_sha256": value.checksum_artifact.sha256,
"checksum_bytes": value.checksum_artifact.bytes,
"checksum_sidecar_path": str(
value.checksum_artifact.manifest_path.relative_to(raw_root)
),
"checksum_sidecar_sha256": value.checksum_artifact.manifest_sha256,
"official_zip_sha256": value.upstream_sha256,
"declared_uncompressed_bytes": value.declared_uncompressed_bytes,
}
for (symbol, day), value in sorted(archives.items())
],
}
destination = data_root / "raw_manifest.json"
if destination.exists():
raise ExploratoryStudyError("raw manifest target already exists")
_write_json(destination, payload)
return destination, sha256_file(destination)
def _persist_locks(
selections: Mapping[str, LockedSelection],
stage: Path,
*,
config: M8StudyConfig,
raw_manifest_sha256: str,
source: Mapping[str, object],
) -> tuple[Path, str]:
child_claims: dict[str, dict[str, object]] = {}
for symbol, selection in sorted(selections.items()):
child = stage / "analysis" / "locks" / f"{symbol}.selection.json"
_atomic_write(child, selection.lock.payload_json.encode("utf-8") + b"\n")
child_claims[symbol] = {
"path": str(child.relative_to(stage)),
"sha256": sha256_file(child),
"selection_lock_sha256": selection.lock.sha256,
"fitted_state_sha256": selection.fitted_state.sha256,
"selected_model": selection.selected_model,
}
comparison = stage / "models" / f"{symbol}.validation_candidates.parquet"
comparison.parent.mkdir(parents=True, exist_ok=True)
selection.validation_comparison.write_parquet(comparison, compression="zstd")
payload = {
"schema_version": "exploratory-aggtrades-analysis-lock-v1",
"created_at_utc": utc_now_iso(),
"config_sha256": config.hash,
"config_source_sha256": config.source_sha256,
"protocol_sha256": sha256_file(_protocol_path(config)),
"raw_manifest_sha256": raw_manifest_sha256,
"source": dict(source),
"children": child_claims,
"development_dates": ["2026-08-05", "2026-08-06"],
"heldout_dates": ["2026-08-07", "2026-08-08"],
"heldout_csv_members_opened_before_lock": False,
}
aggregate = stage / "analysis" / "analysis_lock.json"
_write_json(aggregate, payload)
digest = sha256_file(aggregate)
_atomic_write(aggregate.with_suffix(".sha256"), f"{digest} {aggregate.name}\n".encode())
return aggregate, digest
def _lock_guard(
*,
aggregate_path: Path,
aggregate_sha256: str,
config: M8StudyConfig,
project_root: Path,
source: Mapping[str, object],
) -> None:
if sha256_file(config.path) != config.source_sha256:
raise ExploratoryStudyError("configuration changed before held-out member open")
if sha256_file(_protocol_path(config)) != cast(
str, read_json(aggregate_path)["protocol_sha256"]
):
raise ExploratoryStudyError("protocol changed before held-out member open")
if sha256_file(aggregate_path) != aggregate_sha256:
raise ExploratoryStudyError("aggregate lock changed before held-out member open")
if _source_authority(project_root) != dict(source):
raise ExploratoryStudyError("source identity changed before held-out member open")
aggregate = _mapping(read_json(aggregate_path), "analysis lock")
children = _mapping(aggregate["children"], "analysis lock children")
for claim in children.values():
child = _mapping(claim, "child lock claim")
path = aggregate_path.parents[1] / str(child["path"])
if sha256_file(path) != child["sha256"]:
raise ExploratoryStudyError("child lock changed before held-out member open")
def _checksums(root: Path) -> tuple[Path, str]:
paths = sorted(
item
for item in root.rglob("*")
if item.is_file() and item.name not in {"CHECKSUMS.sha256", "_SUCCESS"}
)
lines = [f"{sha256_file(item)} {item.relative_to(root).as_posix()}" for item in paths]
destination = root / "CHECKSUMS.sha256"
_atomic_write(destination, ("\n".join(lines) + "\n").encode("utf-8"))
return destination, sha256_file(destination)
def verify_exploratory_run(run_dir: str | Path) -> dict[str, object]:
root = Path(run_dir).resolve()
marker = root / "_SUCCESS"
if marker.read_bytes() != SUCCESS_BYTES:
raise ExploratoryStudyError("run success marker is missing or invalid")
checksums = root / "CHECKSUMS.sha256"
for line in checksums.read_text(encoding="utf-8").splitlines():
digest, relative = line.split(" ", 1)
path = root / relative
if not path.is_file() or path.is_symlink() or sha256_file(path) != digest:
raise ExploratoryStudyError(f"run artifact failed checksum verification: {relative}")
evidence = _mapping(read_json(root / "data" / "input_evidence.json"), "input evidence")
for item in cast(Sequence[object], evidence["files"]):
claim = _mapping(item, "input evidence file")
path = Path(str(claim["absolute_path"]))
if path.stat().st_size != int(claim["bytes"]) or sha256_file(path) != claim["sha256"]:
raise ExploratoryStudyError(f"external input changed: {path}")
manifest = root / "run_manifest.json"
return {
"status": "COMPLETE",
"integrity": "verified",
"run_dir": str(root),
"run_manifest": str(manifest),
"run_manifest_sha256": sha256_file(manifest),
"checksums": str(checksums),
"checksums_sha256": sha256_file(checksums),
}
def run_exploratory_study(
config_path: str | Path,
data_root: str | Path,
run_dir: str | Path,
) -> dict[str, object]:
config = _load_config(Path(config_path))
project_root = config.path.parent.parent.resolve()
source = _source_authority(project_root)
destination_data = Path(data_root).resolve()
destination_run = Path(run_dir).resolve()
if destination_run.exists():
raise ExploratoryStudyError("run target already exists; immutable runs are not overwritten")
if (destination_data / "derived").exists() or (destination_data / "raw_manifest.json").exists():
raise ExploratoryStudyError("derived/input-manifest target already exists")
destination_data.mkdir(parents=True, exist_ok=True)
raw_root = destination_data / "raw"
budget = RetainedEvidenceBudget(raw_root, config.study.max_total_download_bytes)
metadata_client = BinancePublicClient(retained_evidence_budget=budget)
archive_client = BinanceArchiveClient(retained_evidence_budget=budget)
metadata: dict[str, M8SymbolMetadata] = {}
raw_metadata: dict[str, SymbolMetadata] = {}
for symbol in config.study.symbols:
observed = metadata_client.fetch_exchange_info(symbol=symbol, raw_root=raw_root)
raw_metadata[symbol] = observed
metadata[symbol] = _metadata_authority(observed)
limits = ArchiveDownloadLimits(
max_compressed_bytes=config.study.max_archive_compressed_bytes,
max_uncompressed_bytes=config.study.max_archive_uncompressed_bytes,
)
archives: dict[tuple[str, str], AcquiredDailyArchive] = {}
for period in config.periods:
for symbol in config.study.symbols:
item = raw_metadata[symbol]
archives[(symbol, period.date.isoformat())] = archive_client.acquire(
DailyArchiveRequest(
symbol=symbol,
date=period.date,
tick_size=item.tick_size,
lot_size=item.lot_size,
),
raw_root=raw_root,
limits=limits,
)
raw_manifest_path, raw_manifest_sha256 = _raw_manifest(
config, metadata, archives, destination_data
)
if sha256_file(config.path) != config.source_sha256:
raise ExploratoryStudyError("configuration changed during raw acquisition")
stage = destination_run.parent / f".{destination_run.name}.staging-{os.getpid()}"
if stage.exists():
raise ExploratoryStudyError("run staging directory already exists")
stage.mkdir(parents=True)
normalized: dict[tuple[str, str], M8ArchiveEntry] = {}
evaluation_paths: dict[tuple[str, str], Path] = {}
selections: dict[str, LockedSelection] = {}
try:
development = config.periods[:2]
heldout = config.periods[2:]
for period in development:
for symbol in config.study.symbols:
normalized_result = normalize_m8_archive(
config,
period,
metadata[symbol],
archives[(symbol, period.date.isoformat())],
raw_root,
output_root=destination_data / "derived",
)
normalized[(symbol, period.date.isoformat())] = normalized_result.entry
evaluation_paths[(symbol, period.date.isoformat())] = _build_evaluation_frame(
normalized_result.entry, config, destination_data
)
for symbol_index, symbol in enumerate(config.study.symbols):
frames = [
pl.read_parquet(evaluation_paths[(symbol, period.date.isoformat())])
for period in development
]
selections[symbol] = select_multidate_model(
frames,
_model_config(config),
feature_columns=_feature_columns(config),
declared_test_dates=[item.date.isoformat() for item in heldout],
seed=config.study.seed + symbol_index,
calibration_bins=config.study.feature_stability_bins,
target=config.study.target,
calibration_fraction=config.study.calibration_fraction,
bootstrap_draws=config.study.bootstrap_samples,
block_width_events=config.study.bootstrap_block_events,
)
del frames
aggregate_path, aggregate_sha256 = _persist_locks(
selections,
stage,
config=config,
raw_manifest_sha256=raw_manifest_sha256,
source=source,
)
for period in heldout:
for symbol in config.study.symbols:
normalized_result = normalize_m8_archive(
config,
period,
metadata[symbol],
archives[(symbol, period.date.isoformat())],
raw_root,
output_root=destination_data / "derived",
before_member_open=lambda: _lock_guard(
aggregate_path=aggregate_path,
aggregate_sha256=aggregate_sha256,
config=config,
project_root=project_root,
source=source,
),
)
normalized[(symbol, period.date.isoformat())] = normalized_result.entry
evaluation_paths[(symbol, period.date.isoformat())] = _build_evaluation_frame(
normalized_result.entry, config, destination_data
)
summaries: dict[str, object] = {}
for symbol in config.study.symbols:
development_frames = [
pl.read_parquet(evaluation_paths[(symbol, period.date.isoformat())])
for period in development
]
test_frames = [
pl.read_parquet(evaluation_paths[(symbol, period.date.isoformat())])
for period in heldout
]
locked = AnalysisLock.restore(
selections[symbol].lock.payload_json,
selections[symbol].lock.sha256,
)
evaluation_result = evaluate_locked_multidate_tests(
development_frames, test_frames, locked
)
model_root = stage / "models" / symbol
metric_root = stage / "metrics" / symbol
model_root.mkdir(parents=True, exist_ok=True)
metric_root.mkdir(parents=True, exist_ok=True)
evaluation_result.predictions.write_parquet(
model_root / "selected_and_prior_predictions.parquet", compression="zstd"
)
evaluation_result.paired_log_loss.per_date.write_parquet(
metric_root / "paired_log_loss_by_date.parquet", compression="zstd"
)
evaluation_result.feature_stability.write_parquet(
metric_root / "feature_stability.parquet", compression="zstd"
)
aggregate = evaluation_result.paired_log_loss.aggregate
summaries[symbol] = {
"selected_model": evaluation_result.selected_model,
"selection_lock_sha256": evaluation_result.lock_sha256,
"replication_status": evaluation_result.paired_log_loss.replication_status,
"equal_date_selected_minus_prior_log_loss": {
"point_estimate": aggregate.point_estimate,
"ci_low": aggregate.lower,
"ci_high": aggregate.upper,
"bootstrap_samples": aggregate.n_bootstrap,
"blocks": aggregate.n_blocks,
"status": aggregate.status,
},
"per_date": evaluation_result.paired_log_loss.per_date.to_dicts(),
}
del development_frames, test_frames, evaluation_result
evidence_files: list[dict[str, object]] = []
for path in sorted(
item for item in destination_data.rglob("*") if item.is_file() and not item.is_symlink()
):
evidence_files.append(
{
"absolute_path": str(path),
"sha256": sha256_file(path),
"bytes": path.stat().st_size,
}
)
_write_json(
stage / "data" / "input_evidence.json",
{
"raw_manifest": str(raw_manifest_path),
"raw_manifest_sha256": raw_manifest_sha256,
"files": evidence_files,
},
)
archive_rows = [
_entry_payload(normalized[(symbol, period.date.isoformat())], destination_data)
for period in config.periods
for symbol in config.study.symbols
]
_write_json(stage / "metrics" / "exploratory_summary.json", summaries)
_write_json(
stage / "run_manifest.json",
{
"schema_version": SCHEMA_VERSION,
"status": "COMPLETE",
"evidence_tier": EVIDENCE_TIER,
"evidence_scope": "trade_only_complete_public_daily_archives_retrospective",
"generated_at_utc": utc_now_iso(),
"config": config.public_dict(),
"protocol": {
"path": PROTOCOL_RELATIVE_PATH,
"sha256": sha256_file(_protocol_path(config)),
},
"source": source,
"raw_manifest_sha256": raw_manifest_sha256,
"analysis_lock_sha256": aggregate_sha256,
"archive_evidence": archive_rows,
"symbol_results": summaries,
"claims": {
"p_values_computed": False,
"significance_claim_authorized": False,
"cross_instrument_pooling": False,
"execution": "NOT_RUN",
"profitability_claim_authorized": False,
"live_l2_claim_authorized": False,
},
},
)
report_lines = [
"# August 5-8 aggregate-trade exploratory result",
"",
"**PUBLIC_ARCHIVE_EXPLORATORY — retrospective trade-only evidence.**",
"",
"This result contains no L2, execution, profitability, capacity, or significance claim.",
"",
]
for symbol in config.study.symbols:
summary = cast(Mapping[str, Any], summaries[symbol])
report_aggregate = cast(
Mapping[str, Any], summary["equal_date_selected_minus_prior_log_loss"]
)
report_lines.extend(
[
f"## {symbol}",
"",
f"- Selected model: `{summary['selected_model']}`",
f"- Equal-date selected-minus-prior log-loss delta: `{report_aggregate['point_estimate']}`",
f"- Descriptive 95% interval: `[{report_aggregate['ci_low']}, {report_aggregate['ci_high']}]`",
f"- Directional replication status: `{summary['replication_status']}`",
"",
]
)
_atomic_write(stage / "reports" / "technical_report.md", "\n".join(report_lines).encode())
_checksums(stage)
_atomic_write(stage / "_SUCCESS", SUCCESS_BYTES)
destination_run.parent.mkdir(parents=True, exist_ok=True)
os.rename(stage, destination_run)
parent = os.open(destination_run.parent, os.O_RDONLY)
try:
os.fsync(parent)
finally:
os.close(parent)
except BaseException:
if stage.exists():
shutil.rmtree(stage)
raise
return verify_exploratory_run(destination_run)
def _parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--config", required=True, type=Path)
parser.add_argument("--data-root", required=True, type=Path)
parser.add_argument("--run-dir", required=True, type=Path)
parser.add_argument("--verify-only", action="store_true")
return parser
def main() -> int:
args = _parser().parse_args()
result = (
verify_exploratory_run(args.run_dir)
if args.verify_only
else run_exploratory_study(args.config, args.data_root, args.run_dir)
)
print(json.dumps(result, sort_keys=True, allow_nan=False))
return 0
if __name__ == "__main__":
raise SystemExit(main())
|