Datasets:
Tasks:
Tabular Classification
Formats:
parquet
Languages:
English
Size:
< 1K
Tags:
economics
quantitative-finance
causal-inference
macroeconomics
housing-economics
market-microstructure
License:
File size: 6,574 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 | from __future__ import annotations
from datetime import datetime
import polars as pl
from microstructure.research.l2_analysis import build_l2_descriptive_analysis
SECOND = 1_000_000_000
DATES = (
("2026-08-10", "train"),
("2026-08-11", "validation"),
("2026-08-12", "primary_test"),
("2026-08-13", "replication_test"),
)
ENDPOINTS = (
("event_20", "event", 20, "events"),
("event_100", "event", 100, "events"),
("clock_1000ms", "clock", 1_000, "milliseconds"),
("clock_5000ms", "clock", 5_000, "milliseconds"),
)
FEATURES = (
"spread_bps",
"depth_total_l1",
"ofi_w20",
"realized_volatility_w100",
)
def _frame(
study_date: str,
study_role: str,
symbol: str,
endpoint_name: str,
domain: str,
horizon: int,
unit: str,
) -> pl.DataFrame:
start = int(datetime.fromisoformat(f"{study_date}T14:00:00+00:00").timestamp()) * SECOND
continuity = f"{study_date}::{symbol}::observed-0000"
rows: list[dict[str, object]] = []
for sequence in range(140):
decision = start + sequence * SECOND
censored = sequence >= 120
label_end = decision + (horizon * SECOND if domain == "event" else horizon * 1_000_000)
ofi = float((sequence % 9) - 4)
future_return = ofi * 1e-5 + (0.25e-5 if symbol == "BTCUSDT" else -0.1e-5)
bid_quantity = 1.0 if sequence % 19 == 0 else 5.0 + (sequence % 7) * 0.2
ask_quantity = 1.2 if sequence % 23 == 0 else 4.5 + (sequence % 5) * 0.2
spread = 8.0 if sequence % 17 == 0 else 1.0 + (sequence % 3) * 0.1
rows.append(
{
"study_date": study_date,
"study_role": study_role,
"endpoint_name": endpoint_name,
"endpoint_domain": domain,
"endpoint_horizon_value": horizon,
"endpoint_horizon_unit": unit,
"symbol": symbol,
"continuity_id": continuity,
"observed_interval_id": continuity,
"observed_interval_start_ns": start,
"observed_interval_end_ns_exclusive": start + 3_600 * SECOND,
"decision_ts_ns": decision,
"decision_sequence": sequence,
"feature_cutoff_ts_ns": decision,
"max_feature_source_ts_ns": decision,
"max_feature_source_sequence": sequence,
"feature_continuity_id": continuity,
"label_start_ts_ns": decision,
"label_start_sequence": sequence,
"right_censored": censored,
"future_mid_return": None if censored else future_return,
"future_mid_up": None if censored else int(future_return > 0),
"label_information_end_ts_ns": None if censored else label_end,
"label_information_end_sequence": None if censored else sequence + horizon,
"label_continuity_id": None if censored else continuity,
"ofi_signed_future_mid_markout_bps": (
None
if censored
else (1.0 if ofi > 0 else -1.0 if ofi < 0 else 0.0) * future_return * 10_000.0
),
"signed_markout_side_source": (
"ofi_w20" if endpoint_name in {"event_20", "clock_1000ms"} else "ofi_w100"
),
"sample_id": f"{study_date}::{symbol}::{endpoint_name}::{sequence}",
"mid_price": 100.0 + sequence * 0.01,
"bid_quantity": bid_quantity,
"ask_quantity": ask_quantity,
"spread_bps": spread,
"depth_total_l1": bid_quantity + ask_quantity,
"depth_total_l5": bid_quantity + ask_quantity + 10.0,
"depth_total_l10": bid_quantity + ask_quantity + 20.0,
"queue_imbalance_l1": (bid_quantity - ask_quantity) / (bid_quantity + ask_quantity),
"realized_volatility_w100": 0.001 + (sequence % 13) * 0.0001,
"ofi_w20": ofi,
"ofi_w100": ofi * 0.8,
"volatility_regime": ("low" if sequence % 3 == 0 else "high"),
"liquidity_regime": ("liquid" if sequence % 4 else "stressed"),
}
)
return pl.DataFrame(rows, infer_schema_length=None)
def _all_frames() -> list[pl.DataFrame]:
return [
_frame(study_date, role, symbol, name, domain, horizon, unit)
for study_date, role in DATES
for symbol in ("BTCUSDT", "ETHUSDT")
for name, domain, horizon, unit in ENDPOINTS
]
def test_l2_descriptive_outputs_are_date_symbol_endpoint_explicit() -> None:
result = build_l2_descriptive_analysis(
_all_frames(), feature_columns=FEATURES, stability_bins=5
)
assert result.intraday_liquidity.height == 8 * 3
assert result.ofi_return_association.height == 32
assert result.signal_half_life.height == 32
assert result.liquidity_recovery.height == 16
assert result.regime_diagnostics.height > 32
assert result.feature_stability.height == 2 * 2 * 4 * len(FEATURES)
assert result.cross_instrument_stability.height == 16
assert result.cross_instrument_stability.get_column("cross_instrument_pooling").not_().all()
assert set(result.ofi_return_association.get_column("interpretation").unique()) == {
"descriptive_book_flow_markout_not_trade_impact"
}
def test_shock_thresholds_and_stability_reference_are_development_only() -> None:
frames = _all_frames()
original = build_l2_descriptive_analysis(frames, feature_columns=FEATURES, stability_bins=5)
mutated = [
frame.with_columns(
pl.when(pl.col("study_role").is_in(["primary_test", "replication_test"]))
.then(pl.col("spread_bps") * 100.0)
.otherwise(pl.col("spread_bps"))
.alias("spread_bps")
)
for frame in frames
]
changed = build_l2_descriptive_analysis(mutated, feature_columns=FEATURES, stability_bins=5)
assert (
original.liquidity_recovery.select(
"symbol", "train_spread_q95", "train_executable_depth_q05"
)
.unique()
.sort("symbol")
.equals(
changed.liquidity_recovery.select(
"symbol", "train_spread_q95", "train_executable_depth_q05"
)
.unique()
.sort("symbol")
)
)
assert set(changed.feature_stability.get_column("reference_scope").unique()) == {
"train_plus_validation_only"
}
|