Decomposition-First Quant Trading Tutorials with DeTime

This tutorial sequence rebuilds the quant-trading examples around one thesis: classic technical strategies are rough, implicit estimates of trend, cycle, residual deviation, and market participation. DeTime makes those structures explicit before the strategy layer.

The revised tutorial uses a seven-part path rather than a loose collection of indicator notebooks.

Open the executed tutorial notebook index to jump directly between Tutorial 00-07, the strategy lab notebooks, and the strategy expansion notebooks.

Tutorial	Status	Topic	Main DeTime role
00	implemented	roadmap	strategy design map
01	implemented	market data and feature factory	price + volume decomposition
02	implemented	moving average, multi-MA and MACD	explicit trend filters with cycle/residual/volume gates
03	implemented	RSI, Bollinger and residual mean reversion	residual deviation, cycle timing and volume filter
04	implemented	Turtle/Donchian breakout	trend, cycle, residual overextension and volume confirmation
05	implemented	pairs trading and stat arb	spread trend drift + spread residual + optional volume news filter
06	implemented	cross-sectional rotation	trend/cycle/residual/volume decomposition factor score
07	implemented	native SSA high-return / low-drawdown replay	fast/slow SSA trends, next-open fills, drawdown and monthly trade diagnostics

Strategy-lab correction: two concrete trading systems

The clearest entry point is now the strategy lab, not the broad seven-tutorial map. It implements two complete strategy families:

1. Trend following: trend_slope and trend_strength create the trading signal; cycle_position, residual_abs_z, and volume decomposition control entry timing and overextension.
2. Oscillation / residual reversion: when abs(trend_strength) is small, the strategy trades residual_z around trend + cycle; negative residual buys, positive residual sells or shorts.

Run it with:

make strategy-lab

Main implementation files:

examples/quant_trading/strategy_lab.py
examples/quant_trading/scripts/run_strategy_lab.py
examples/notebooks/quant_trading/01_detime_trend_following_strategy_lab.ipynb
examples/notebooks/quant_trading/02_detime_oscillation_reversion_strategy_lab.ipynb
docs/tutorials/quant-trading/two-strategy-families.md

The script writes strategy stats, order records, round-trip trades, feature coverage, run manifest, and buy/sell charts under examples/quant_trading/reports/strategy_lab/.

Implemented notebooks

The executable notebooks are in examples/notebooks/quant_trading/: install the optional notebook and market-data dependencies from examples/quant_trading/requirements.txt. The rendered documentation publishes the captured outputs directly at quant-trading/notebooks/01_market_data_and_decomposition_feature_factory.md and the companion pages in the same notebook directory.

Notebook	What it teaches
00_decomposition_first_quant_trading_roadmap.ipynb	Why the tutorial is organized around decomposition rather than isolated indicators.
01_market_data_and_decomposition_feature_factory.ipynb	OHLCV audit, period estimation, and walk-forward price/volume feature construction.
02_decomposition_aware_moving_average_macd.ipynb	Classical buy-and-hold, dual MA, MACD, multi-MA and momentum compared with DeTime rewrites.
03_residual_mean_reversion_rsi_bollinger.ipynb	Price z-score, RSI, Bollinger and APO baselines rewritten as residual mean-reversion with cycle timing.
04_turtle_donchian_breakout_volume_confirmation.ipynb	Donchian/Turtle breakout rewritten with trend, cycle, residual and volume confirmation.
05_pairs_spread_decomposition_stat_arb.ipynb	Pair z-score and rolling-beta spread baselines rewritten as residual spread trading with spread-trend drift control.
06_cross_sectional_rotation_portfolio.ipynb	Momentum, multi-MA and inverse-volatility rotation compared with DeTime cross-sectional scoring.
07_native_ssa_high_return_low_drawdown_tutorial.ipynb	Native SSA dual-trend mean-reversion replay for selected high-return FX and crypto strategies under a 20% drawdown constraint.

Core design

For price and volume we use:

[ \log P_t = T^P_t + C^P_t + R^P_t, \qquad \log(1 + V_t) = T^V_t + C^V_t + R^V_t. ]

The strategy layer then reads:

Component	Trading interpretation	Example use
price trend	direction and persistence	trend-following state, decomposed MA/MACD
price cycle	timing and local oscillation	avoid buying into overextended cycle peaks
price residual	deviation from current structure	pullback and mean-reversion logic
volume trend	participation	confirm trend or breakout
volume residual	abnormal activity	detect volume shock or weak participation
reconstruction error / stability	feature reliability	reduce exposure when component quality weakens

Data layer

The code supports live downloads through yfinance. The rendered documentation uses archived historical GOOG and FX OHLCV samples from the user-provided Learn Algorithmic Trading material so the pages can be rebuilt without network access.

For formal trading research, replace the educational data source with a licensed point-in-time vendor and document symbol membership, corporate actions, delistings, borrow, funding, FX, and execution assumptions.

Code modules

Module	Purpose
data.py	OHLCV download, panel extraction, archived market-data loader
classic_indicators.py	SMA, EMA, MACD, RSI, Bollinger, momentum, APO
decomposition_features.py	walk-forward price and volume DeTime features
strategy_baselines.py	classical baseline weight recipes
strategy_detime.py	decomposition-aware strategy recipes
strategy_pairs.py	pair spread decomposition and stat-arb recipes
strategy_rotation.py	cross-sectional rotation and portfolio recipes
validation.py	common backtest comparison, turnover, manifest helpers
backtest.py	transparent close-to-close research backtester

Smoke test

export PYTHONPATH="$PWD/src:$PWD/examples:$PYTHONPATH"
make smoke-05-06

The smoke tests run on CPU and write hardware/audit outputs under examples/quant_trading/reports/.

Latest Implemented Tutorials

Tutorials 03-04 cover single-asset reversion and breakout strategies:

• Tutorial 03 keeps familiar entry points, including RSI, Bollinger Bands, APO and price z-score, but changes the traded object from raw price deviation to residual deviation after trend/cycle removal.
• Tutorial 04 keeps the Donchian/Turtle breakout scaffold but adds trend, cycle, residual-overextension and volume-participation gates.

Tutorials 05-07 complete the current arc:

• Tutorial 05 rewrites pairs trading by decomposing the rolling hedge spread and trading residual deviation only when spread trend drift is controlled.
• Tutorial 06 turns decomposition outputs into a cross-sectional factor score for top-N, long-short and volatility-targeted rotation portfolios.
• Tutorial 07 replays selected native SSA dual-trend strategies, including buy/sell markers, strategy-vs-buy-hold equity, drawdown curves, and monthly return / win-rate tables.

Run the latest batch on bundled real FX samples:

make quant-columns-05-06

Run the bundled-sample Tutorial 01-06 batch:

make quant-columns-01-06

Tutorial 07 is a notebook replay for local 3-minute intraday files. Open it from the rendered documentation or run examples/notebooks/quant_trading/07_native_ssa_high_return_low_drawdown_tutorial.ipynb after placing the referenced FX and crypto CSVs under examples/quant_trading/data/intraday_crypto_fx/.

Live-data versions are available through make quant-columns-03-04-live and make quant-columns-05-06-live.

Strategy expansion: method-specific variants and component pair trading

The strategy lab now includes an additional expansion layer that treats a decomposition method and horizon as part of the strategy definition.

make strategy-expansion

This command runs:

1. method-specific strategy variants: STL/SSA/STD configurations produce different trend, cycle, and residual inputs for the same trend-following, oscillation-reversion, residual-Bollinger, trend-MACD, and trend-crossover rules;
2. component-level pair trading: both assets in a pair are decomposed, trend/cycle similarity is measured, cointegration and stationarity diagnostics are reported, and the residual gap or trend+cycle spread deviation is traded.

Tutorial document:

docs/tutorials/quant-trading/method-specific-strategy-expansion.md

New notebooks:

examples/notebooks/quant_trading/03_detime_method_specific_strategy_variants.ipynb
examples/notebooks/quant_trading/04_detime_component_pair_trading_cointegration.ipynb