Tutorial 01 - Real market data and the DeTime feature factory

from pathlib import Path

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from IPython.display import display

from examples.quant_trading.data import (
    load_sample_goog_ohlcv,
    market_data_manifest,
    ohlcv_audit_report,
)
from examples.quant_trading.decomposition_features import (
    build_feature_table,
    estimate_dominant_period,
    feature_coverage_report,
)
from examples.quant_trading.features import decompose_one_series, walkforward_decompose_ohlcv
from examples.quant_trading.validation import write_run_audit

pd.set_option("display.max_columns", 20)
REPORT_DIR = Path("examples/quant_trading/reports")
REPORT_DIR.mkdir(parents=True, exist_ok=True)

ohlcv_single = load_sample_goog_ohlcv(trim_start="2014-01-01")
ticker = ohlcv_single.attrs.get("symbol", "GOOG")
ohlcv = {
    field: ohlcv_single[[field]].rename(columns={field: ticker})
    for field in ["Open", "High", "Low", "Close", "Volume"]
}
close = ohlcv["Close"]
volume = ohlcv["Volume"]

audit = ohlcv_audit_report(ohlcv)
manifest = market_data_manifest(
    tickers=[ticker],
    start=str(close.index.min().date()),
    end=str(close.index.max().date()),
    interval="1d",
    source=ohlcv_single.attrs.get("source", "bundled historical OHLCV sample"),
)

display(audit)
display(manifest)

fig, ax1 = plt.subplots(figsize=(10, 4))
close[ticker].plot(ax=ax1, linewidth=1.4, label="close")
ax1.set_title("Historical close and volume")
ax1.set_ylabel("Close")
ax2 = ax1.twinx()
volume[ticker].plot(ax=ax2, alpha=0.25, linewidth=0.8, label="volume")
ax2.set_ylabel("Volume")
ax1.grid(True, alpha=0.25)
plt.show()

image/png

period_estimate = estimate_dominant_period(close[ticker], candidates=(63, 126, 252), use_log=True)
period_summary = pd.DataFrame([period_estimate.__dict__])
display(period_summary)

features = walkforward_decompose_ohlcv(
    ohlcv,
    method="STL",
    period="auto",
    period_candidates=(63, 126, 252),
    train_window=504,
    step=5,
    z_window=63,
)
coverage = feature_coverage_report(features)
display(coverage.sort_values(["feature", "asset"]).head(18))

feature_table = build_feature_table(close, features)
latest = feature_table.tail(5)
display(latest)

diagnostic = decompose_one_series(
    close[ticker],
    method="STL",
    period=int(period_estimate.period),
    z_window=63,
    transform="log",
)
trend_price = np.exp(diagnostic["trend"])
fair_value = np.exp(diagnostic["trend"] + diagnostic["cycle"])

fig, ax = plt.subplots(figsize=(10, 4))
close[ticker].plot(ax=ax, linewidth=1.0, color="#1f2937", label="close")
trend_price.plot(ax=ax, linewidth=2.0, color="#0f766e", label="continuous DeTime trend")
fair_value.plot(ax=ax, linewidth=1.3, color="#7c3aed", alpha=0.85, label="trend + cycle fair value")
ax.set_title("Continuous diagnostic decomposition: trend and trend + cycle")
ax.legend()
ax.grid(True, alpha=0.25)
plt.show()

image/png

fig, ax = plt.subplots(figsize=(10, 3))
diagnostic["cycle"].plot(ax=ax, linewidth=1.2, color="#2563eb")
ax.axhline(0, linewidth=0.8, color="#64748b")
ax.set_title("Cycle component: timing context around the trend")
ax.grid(True, alpha=0.25)
plt.show()

image/png

fig, ax = plt.subplots(figsize=(10, 3))
diagnostic["residual_z"].plot(ax=ax, linewidth=1.2, color="#dc2626")
ax.axhline(2.0, linestyle="--", linewidth=0.9, color="#991b1b")
ax.axhline(-2.0, linestyle="--", linewidth=0.9, color="#991b1b")
ax.axhline(0, linewidth=0.8, color="#64748b")
ax.set_title("Residual pressure after trend and cycle removal")
ax.grid(True, alpha=0.25)
plt.show()

image/png

fig, ax = plt.subplots(figsize=(10, 3))
features["trend_strength"][ticker].plot(ax=ax, linewidth=1.2, label="walk-forward trend strength")
features["component_stability"][ticker].plot(ax=ax, linewidth=1.2, label="walk-forward component stability")
ax.set_title("Causal walk-forward diagnostics emitted by the feature factory")
ax.legend()
ax.grid(True, alpha=0.25)
plt.show()

image/png

fig, ax = plt.subplots(figsize=(10, 3))
features["volume_trend_slope"][ticker].plot(ax=ax, linewidth=1.1, label="volume trend slope")
features["volume_residual_z"][ticker].plot(ax=ax, linewidth=1.0, label="volume residual z")
ax.axhline(0, linewidth=0.8)
ax.set_title("Volume decomposition features")
ax.legend()
ax.grid(True, alpha=0.25)
plt.show()

image/png

fig, ax = plt.subplots(figsize=(10, 2.8))
features["selected_period"][ticker].dropna().plot(ax=ax, drawstyle="steps-post")
ax.set_title("Selected period over walk-forward windows")
ax.set_ylabel("trading days")
ax.grid(True, alpha=0.25)
plt.show()

image/png

feature_table.tail(60).to_csv(REPORT_DIR / "column_01_feature_table_tail.csv")
paths = write_run_audit(
    REPORT_DIR,
    data_manifest=manifest,
    audit=audit,
    strategy_stats=None,
    prefix="column_01",
)
summary = pd.DataFrame({"artifact": list(paths), "path": [str(p) for p in paths.values()]})
display(summary)