Tutorial 06 - Cross-sectional rotation and portfolio construction

Executed tutorial notebook. This page is generated from examples/notebooks/quant_trading/06_cross_sectional_rotation_portfolio.ipynb and includes markdown cells, code cells, stdout, tables, and captured figures from the committed notebook.

Tutorial Navigation

Track Tutorial notebook
Roadmap Tutorial 00 - Roadmap
Strategy Lab 01 Trend-Following Lab
Tutorial Sequence 01 Real Market Data and Feature Factory
Tutorial Sequence 02 Decomposition-aware MA and MACD
Strategy Lab 02 Oscillation-Reversion Lab
Strategy Expansion 03 Method-Specific Variants
Tutorial Sequence 03 Residual Mean Reversion
Strategy Expansion 04 Component Pair Trading
Tutorial Sequence 04 Donchian Breakout
Tutorial Sequence 05 Pair-Spread Stat-Arb
Tutorial Sequence 06 Cross-Sectional Rotation
Native SSA Replay 07 Native SSA High-Return / Low-Drawdown

Executed Notebook

This notebook turns DeTime features into a portfolio ranking language: trend decides direction, cycle adjusts timing, residual controls overextension and volume/reconstruction features control reliability.

In [1]
import matplotlib.pyplot as plt
import pandas as pd

from quant_trading.data import load_bundled_real_ohlcv_panel, ohlcv_audit_report
from quant_trading.features import walkforward_decompose_ohlcv
from quant_trading.decomposition_features import feature_coverage_report, build_feature_table
from quant_trading.strategy_baselines import buy_and_hold_weights
from quant_trading.strategy_rotation import (
    classic_momentum_rotation_weights, detime_cross_sectional_score,
    detime_long_short_rotation_weights, detime_rotation_weights,
    rotation_diagnostic_table, volume_availability,
)
from quant_trading.validation import compare_weight_strategies, turnover_report

1. Load real offline market data

In [2]
tickers = ["AUDUSD=X", "NZDUSD=X", "EURUSD=X", "GBPUSD=X"]
ohlcv = load_bundled_real_ohlcv_panel(tickers, min_observations=120)
ohlcv = {field: table.tail(760).copy() for field, table in ohlcv.items()}
prices = ohlcv["Close"]
volumes = ohlcv.get("Volume")
print("volume_available:", volume_availability(volumes))
ohlcv_audit_report(ohlcv)
stdout
volume_available: False
text/html
ticker first_timestamp last_timestamp observations close_missing_ratio volume_missing_ratio zero_volume_ratio min_close max_close median_volume
0 AUDUSD=X 2015-02-02 2018-01-02 760 0.0 0.0 1.0 0.686106 0.811293 0.0
1 NZDUSD=X 2015-02-02 2018-01-02 760 0.0 0.0 1.0 0.626684 0.772380 0.0
2 EURUSD=X 2015-02-02 2018-01-02 760 0.0 0.0 1.0 1.039047 1.202906 0.0
3 GBPUSD=X 2015-02-02 2018-01-02 760 0.0 0.0 1.0 1.203935 1.588512 0.0

2. Build walk-forward asset features

In [3]
features = walkforward_decompose_ohlcv(
    ohlcv,
    method="STL",
    period="auto",
    period_candidates=(63, 126, 252),
    train_window=504,
    step=5,
    z_window=63,
)
feature_coverage_report(features).query("feature in ['trend_slope', 'trend_strength', 'cycle_slope', 'residual_z', 'residual_abs_z', 'volume_participation']").head(18)
text/html
feature asset observations non_null coverage first_valid last_valid
12 trend_slope AUDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
13 trend_slope NZDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
14 trend_slope EURUSD=X 760 257 0.338158 2017-01-05 2018-01-02
15 trend_slope GBPUSD=X 760 257 0.338158 2017-01-05 2018-01-02
20 trend_strength AUDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
21 trend_strength NZDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
22 trend_strength EURUSD=X 760 257 0.338158 2017-01-05 2018-01-02
23 trend_strength GBPUSD=X 760 257 0.338158 2017-01-05 2018-01-02
32 cycle_slope AUDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
33 cycle_slope NZDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
34 cycle_slope EURUSD=X 760 257 0.338158 2017-01-05 2018-01-02
35 cycle_slope GBPUSD=X 760 257 0.338158 2017-01-05 2018-01-02
48 residual_z AUDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
49 residual_z NZDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
50 residual_z EURUSD=X 760 257 0.338158 2017-01-05 2018-01-02
51 residual_z GBPUSD=X 760 257 0.338158 2017-01-05 2018-01-02
52 residual_abs_z AUDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
53 residual_abs_z NZDUSD=X 760 257 0.338158 2017-01-05 2018-01-02
In [4]
build_feature_table(prices, features).tail(3).iloc[:, :12].round(4)
text/html
component_stability cycle cycle_amplitude
AUDUSD=X EURUSD=X GBPUSD=X NZDUSD=X AUDUSD=X EURUSD=X GBPUSD=X NZDUSD=X AUDUSD=X EURUSD=X GBPUSD=X NZDUSD=X
Date
2017-12-29 0.9957 0.9939 0.9935 0.9935 -0.0227 -0.0174 0.0056 -0.0092 0.0306 0.0198 0.0103 0.0311
2018-01-01 0.9959 0.9943 0.9932 0.9941 -0.0147 -0.0086 0.0114 0.0075 0.0306 0.0187 0.0100 0.0304
2018-01-02 0.9959 0.9943 0.9932 0.9941 -0.0147 -0.0086 0.0114 0.0075 0.0306 0.0187 0.0100 0.0304

3. Inspect the DeTime rotation score

In [5]
score = detime_cross_sectional_score(prices, features)
score.tail(5).round(3)
text/html
AUDUSD=X NZDUSD=X EURUSD=X GBPUSD=X
Date
2017-12-27 1.894 0.294 2.531 1.656
2017-12-28 1.894 0.294 2.531 1.656
2017-12-29 1.894 0.294 2.531 1.656
2018-01-01 1.656 0.644 2.519 1.556
2018-01-02 1.656 0.644 2.519 1.556
In [6]
fig, ax = plt.subplots(figsize=(10, 4))
score.tail(120).plot(ax=ax, linewidth=1.1)
ax.axhline(0, color="black", linewidth=0.8)
ax.set_title("DeTime cross-sectional score on real FX samples")
ax.set_ylabel("Score")
ax.grid(True, alpha=0.25)
plt.tight_layout()
plt.show()
image/png
Notebook output cell 10
In [7]
rotation_diagnostic_table(prices, features, tail=2).round(4)
text/html
date asset score trend_strength cycle_slope residual_z volume_participation
0 2018-01-01 EURUSD=X 2.5188 0.1228 0.0016 -0.4600 1.25
1 2018-01-01 AUDUSD=X 1.6563 0.0496 0.0010 0.2019 1.25
2 2018-01-01 GBPUSD=X 1.5562 0.0393 0.0011 0.1007 1.25
3 2018-01-01 NZDUSD=X 0.6438 -0.0190 0.0036 -0.8123 1.25
4 2018-01-02 EURUSD=X 2.5188 0.1228 0.0016 -0.4600 1.25
5 2018-01-02 AUDUSD=X 1.6563 0.0496 0.0010 0.2019 1.25
6 2018-01-02 GBPUSD=X 1.5562 0.0393 0.0011 0.1007 1.25
7 2018-01-02 NZDUSD=X 0.6438 -0.0190 0.0036 -0.8123 1.25

4. Backtest a compact rotation suite

In [8]
strategies = {
    "buy_hold_equal_weight": buy_and_hold_weights(prices),
    "classic_momentum_63_top": classic_momentum_rotation_weights(prices, lookback=63, top_n=2, rebalance_freq="W-FRI", vol_target=None),
    "detime_rotation_top_trend_cycle_residual_volume": detime_rotation_weights(prices, features, top_n=2, rebalance_freq="W-FRI", vol_target=None),
    "detime_rotation_long_short": detime_long_short_rotation_weights(prices, features, top_n=2, bottom_n=2),
}
comparison, results = compare_weight_strategies(prices, strategies, fee_bps=1.0, slippage_bps=2.0)
comparison[["total_return", "sharpe", "max_drawdown", "average_turnover"]].round(4)
text/html
total_return sharpe max_drawdown average_turnover
strategy
detime_rotation_long_short 0.0052 0.0686 -0.1563 0.0572
detime_rotation_top_trend_cycle_residual_volume -0.0112 0.0095 -0.1596 0.0289
buy_hold_equal_weight -0.0115 -0.0085 -0.1234 0.0000
classic_momentum_63_top -0.0799 -0.2850 -0.1403 0.0461
In [9]
fig, ax = plt.subplots(figsize=(10, 4))
for name, result in results.items():
    result.equity.plot(ax=ax, linewidth=1.2, label=name)
ax.set_title("Rotation equity curves on real FX samples")
ax.set_ylabel("Equity")
ax.legend(loc="best", fontsize=8)
ax.grid(True, alpha=0.25)
plt.tight_layout()
plt.show()
image/png
Notebook output cell 14
In [10]
turnover_report(strategies).round(4)
text/html
average_turnover median_turnover max_turnover average_gross_exposure
strategy
buy_hold_equal_weight 0.0000 0.0000 0.0 1.0000
classic_momentum_63_top 0.0461 0.0000 2.0 0.9158
detime_rotation_top_trend_cycle_residual_volume 0.0289 0.0000 2.0 0.9947
detime_rotation_long_short 0.0572 0.0034 3.0 1.2333

5. Live-data extension

Run run_column_06_cross_sectional_rotation.py without --use-bundled-sample to use sector ETFs and real equity/ETF volume. In the bundled FX sample, raw volume is unavailable, so volume is neutral rather than invented.