Modéliser les coûts de transaction en trading

# Chapter 18: Transaction Costs


# Chapter 18: Transaction Costs

The chapter reframes transaction costs from a backtest adjustment into a workflow constraint that affects factor evaluation, simulation, portfolio construction, risk management, and production monitoring. Readers should care because it establishes the chapter's central claim: many strategies fail not because the forecast is wrong, but because the implementation problem was ignored.

## Learning Objectives

- Identify where transaction costs enter the ML4T workflow, from factor evaluation and backtesting to portfolio construction, risk management, and production monitoring
- Distinguish explicit, implicit, and capacity-related trading costs and map each component to the relevant modeling choice
- Explain why execution costs vary with market regime, intraday liquidity, volatility, and execution urgency
- Choose and calibrate baseline backtest cost models, from spread-based assumptions to linear and square-root impact models, using conservative research defaults when direct execution data is unavailable
- Compare common execution approaches, including TWAP, VWAP, adaptive participation, and Almgren-Chriss-style optimal execution, in terms of impact, timing risk, and signal decay
- Use transaction cost analysis to decompose realized costs, diagnose model misspecification, and recalibrate ex ante assumptions
- Apply break-even turnover, minimum required edge, alpha-to-go, capacity analysis, and precommitted kill criteria to decide whether a strategy remains economically viable after costs

## Sections

### 18.1 Where Costs Enter the ML4T Workflow

This section reframes transaction costs from a backtest adjustment into a workflow constraint that affects factor evaluation, simulation, portfolio construction, risk management, and production monitoring. Readers should care because it establishes the chapter's central claim: many strategies fail not because the forecast is wrong, but because the implementation problem was ignored.

- [`01_cost_taxonomy`](01_cost_taxonomy.ipynb) — This notebook maps the transaction cost landscape across our seven asset classes, comparing exchange fee structures, spread regimes, and the resulting breakeven alpha requirements for each case study. Uses cme_futures, crypto_perps, etfs and 3 more data.
- [`03_market_impact_calibration`](03_market_impact_calibration.ipynb) — This notebook calibrates market impact models using real market data, estimates Kyle's lambda from NASDAQ-100 trade classification data, maps intraday volume profiles, and estimates strategy capacity limits for each asset class. Uses cme_futures, crypto_perps, etfs and 5 more data.

### 18.2 A Cost Taxonomy for Practitioners

This section organizes costs into explicit, implicit, and capacity components and shows that each one demands a different modeling response. Its practical value is that it stops readers from collapsing everything into a vague slippage assumption and instead teaches them how to think about the actual sources of gross-to-net decay.

- [`01_cost_taxonomy`](01_cost_taxonomy.ipynb) — This notebook maps the transaction cost landscape across our seven asset classes, comparing exchange fee structures, spread regimes, and the resulting breakeven alpha requirements for each case study. Uses cme_futures, crypto_perps, etfs and 3 more data.
- [`02_spread_estimation`](02_spread_estimation.ipynb) — This notebook estimates bid-ask spreads from OHLCV data using two classical estimators, validates them against ground-truth microstructure data, and applies them across all seven asset classes. Uses cme_futures, crypto_perps, etfs and 5 more data.
- [`12_commission_slippage_comparison`](12_commission_slippage_comparison.ipynb) — Provides a side-by-side comparison of all 6 commission models and 5 slippage models in ml4t.backtest.models. We compute costs for identical trades across varying sizes, define 4 asset-class cost stacks, and measure the P&L sensitivity and frequency sensitivity of model choice.

### 18.3 The Microstructure Regime Link

This section explains why cost parameters are not stationary and must be conditioned on time of day, volatility, liquidity, and stress. It matters because even a well-designed cost model becomes dangerous when it treats crisis execution like normal execution or assumes that today's spread and depth are stable inputs.

- [`02_spread_estimation`](02_spread_estimation.ipynb) — This notebook estimates bid-ask spreads from OHLCV data using two classical estimators, validates them against ground-truth microstructure data, and applies them across all seven asset classes. Uses cme_futures, crypto_perps, etfs and 5 more data.
- [`03_market_impact_calibration`](03_market_impact_calibration.ipynb) — This notebook calibrates market impact models using real market data, estimates Kyle's lambda from NASDAQ-100 trade classification data, maps intraday volume profiles, and estimates strategy capacity limits for each asset class. Uses cme_futures, crypto_perps, etfs and 5 more data.

### 18.4 Baseline Backtesting Cost Models

This section gives readers a practical ladder of cost models, from spread-only assumptions to linear slippage and square-root impact. It is useful because it offers a concrete modeling toolkit for research backtests while making clear when simple models are acceptable, when they are too optimistic, and why conservative calibration is often the right default.

- [`03_market_impact_calibration`](03_market_impact_calibration.ipynb) — This notebook calibrates market impact models using real market data, estimates Kyle's lambda from NASDAQ-100 trade classification data, maps intraday volume profiles, and estimates strategy capacity limits for each asset class. Uses cme_futures, crypto_perps, etfs and 5 more data.
- [`06_ml4t_execution_demo`](06_ml4t_execution_demo.ipynb) — This notebook demonstrates the ml4t.backtest.execution module for realistic execution cost modeling. The library provides four market impact models: Uses etfs data.

### 18.5 Execution Algorithms as Controls, Not Magic

This section demystifies execution algorithms by presenting TWAP, VWAP, and regime-aware participation as ways to manage trade-offs rather than eliminate costs. Readers should care because it ties execution design back to signal half-life, urgency, and capacity, showing that execution choices are part of strategy design, not an afterthought left to a trading desk.

- [`04_vwap_twap_execution`](04_vwap_twap_execution.ipynb) — This notebook implements the two most common execution benchmarks: Uses synthetic data.
- [`07_ml4t_volume_participation`](07_ml4t_volume_participation.ipynb) — This notebook demonstrates VolumeParticipationLimit from ml4t.backtest.execution for realistic institutional order execution: Uses synthetic data.
- [`08_ml_dynamic_execution`](08_ml_dynamic_execution.ipynb) — This notebook explores how machine learning can be used to dynamically adapt execution strategies based on real-time market conditions. Instead of following a fixed VWAP/TWAP schedule, we use ML to predict optimal execution parameters.

### 18.6 Optimizing Execution with Almgren-Chriss as a Unifying Framework

This section introduces Almgren-Chriss as the cleanest framework for thinking about impact, timing risk, and urgency in one model. Its importance is less in deriving a perfect trading schedule than in giving readers a disciplined way to reason about execution feasibility and to connect portfolio intent with execution reality.

- [`05_almgren_chriss_optimal_execution`](05_almgren_chriss_optimal_execution.ipynb) — This notebook implements the seminal Almgren-Chriss (2001) framework for optimal trade execution. We derive the efficient frontier of execution strategies, compute optimal trajectories, and demonstrate Transaction Cost Analysis (TCA) methodology.

### 18.7 Transaction Cost Analysis and Model Validation

This section turns realized fills into model feedback through implementation shortfall, decomposition, regime-aware benchmarking, and calibration. It matters because it closes the loop: cost assumptions stop being static research inputs and become hypotheses that are tested, decomposed, and revised against live evidence.

- [`01_cost_taxonomy`](01_cost_taxonomy.ipynb) — This notebook maps the transaction cost landscape across our seven asset classes, comparing exchange fee structures, spread regimes, and the resulting breakeven alpha requirements for each case study. Uses cme_futures, crypto_perps, etfs and 3 more data.
- [`10_gross_vs_net_performance`](10_gross_vs_net_performance.ipynb) — This notebook provides a comprehensive framework for analyzing the gap between gross (theoretical) and net (realized) strategy performance. This is the ultimate reality check for any trading strategy.

### 18.8 Designing Practical Cost Guardrails

This section provides decision rules such as break-even turnover, minimum required edge, alpha-to-go, capacity analysis, and kill criteria. Readers should care because this is where the chapter becomes operational: it shows how to decide whether a strategy can be deployed, scaled, modified, or abandoned once trading frictions are treated honestly.

- [`03_market_impact_calibration`](03_market_impact_calibration.ipynb) — This notebook calibrates market impact models using real market data, estimates Kyle's lambda from NASDAQ-100 trade classification data, maps intraday volume profiles, and estimates strategy capacity limits for each asset class. Uses cme_futures, crypto_perps, etfs and 5 more data.
- [`09_frequency_tradeoff`](09_frequency_tradeoff.ipynb) — This notebook demonstrates the critical tradeoff between signal quality and transaction costs at different rebalancing frequencies. Uses synthetic data.
- [`10_gross_vs_net_performance`](10_gross_vs_net_performance.ipynb) — This notebook provides a comprehensive framework for analyzing the gap between gross (theoretical) and net (realized) strategy performance. This is the ultimate reality check for any trading strategy.
- [`11_cost_cliff`](11_cost_cliff.ipynb) — This notebook demonstrates the dramatic impact of transaction costs on intraday trading strategies. What looks like a stellar strategy on a gross basis often becomes unprofitable or marginal after realistic costs.

The cross-case-study cost-survival comparison lives in Chapter 20: see [`20_strategy_synthesis/06_cost_survival`](../20_strategy_synthesis/06_cost_survival.ipynb).

## Running the Notebooks

```bash
# From the repository root
uv run python 18_transaction_costs/<notebook>.py

# Test mode (reduced data via Papermill)
uv run pytest tests/test_chapter_notebooks.py -v -k "18_transaction_costs"
```

## References

- **Albert S. Kyle** (1985). [Continuous Auctions and Insider Trading](https://doi.org/10.2307/1913210). *Econometrica*.
- **Ananth Madhavan** (2002). [Market Microstructure: A Practitioner's Guide](https://www.jstor.org/stable/4480415). *Financial Analysts Journal*.
- **Andrea Frazzini et al.** (2018). [Trading Costs](https://doi.org/10.2139/ssrn.3229719).
- **Anna A. Obizhaeva and Jiang Wang** (2013). [Optimal trading strategy and supply/demand dynamics](https://doi.org/10.1016/j.finmar.2012.09.001). *Journal of Financial Markets*.
- **Bence Toth et al.** (2011). [Anomalous price impact and the critical nature of liquidity in financial markets](https://doi.org/10.1103/PhysRevX.1.021006). *Physical Review X*.
- **Christopher Schwarz et al.** (2022). [The 'Actual Retail Price' of Equity Trades](https://doi.org/10.2139/ssrn.4189239).
- **Damian Eduardo Taranto et al.** (2018). [Linear models for the impact of order flow on prices. I. History dependent impact models](https://doi.org/10.1080/14697688.2017.1395903). *Quantitative Finance*.
- **Dmitriy Muravyev and Neil D. Pearson** (2020). [Option Trading Costs Are Lower than You Think](https://doi.org/10.2139/ssrn.2580548).
- **Emilio Said** (2022). [Market Impact: Empirical Evidence, Theory and Practice](http://arxiv.org/abs/2205.07385).
- **Giuseppe A. Paleologo** (2025). The Elements of Quantitative Investing. *John Wiley & Sons*.
- **Hector Chan** (2022). [Market Impact Decay and Capacity](https://doi.org/10.3905/jpm.2022.1.382). *The Journal of Portfolio Management*.
- **James O'Donovan and Gloria Yang Yu** (2024). [A Transaction Cost Perspective on Option Anomalies](https://doi.org/10.2139/ssrn.4806038).
- **Jean-Philippe Bouchaud** (2022). [The inelastic market hypothesis: a microstructural interpretation](https://doi.org/10.1080/14697688.2022.2068052). *Quantitative Finance*.
- **Joel Hasbrouck** (1991). [Measuring the Information Content of Stock Trades](https://doi.org/10.2307/2328693). *The Journal of Finance*.
- **Marco Avellaneda and Sasha Stoikov** (2008). [High-frequency trading in a limit order book](https://doi.org/10.1080/14697680701381228). *Quantitative Finance*.
- **Nikolaus Hautsch and Ruihong Huang** (2012). [The market impact of a limit order](https://doi.org/10.1016/j.jedc.2011.09.012). *Journal of Economic Dynamics and Control*.
- **Nina Karnaukh et al.** (2015). [Understanding FX Liquidity](https://doi.org/10.2139/ssrn.2329738).
- **Rama Cont et al.** (2014). [The Price Impact of Order Book Events](https://doi.org/10.1093/jjfinec/nbt003). *Journal of Financial Econometrics*.
- **Robert Almgren and Neil Chriss** (2001). [Optimal execution of portfolio transactions](https://doi.org/10.21314/jor.2001.041). *The Journal of Risk*.
- **Ryan Donnelly** (2022). [Optimal Execution: A Review](https://doi.org/10.1080/1350486X.2022.2161588). *Applied Mathematical Finance*.
- **Steven L. Heston et al.** (2023). [Option Momentum](https://papers.ssrn.com/abstract=4113680).
- **Tarun Chordia et al.** (2000). [Commonality in liquidity](https://doi.org/10.1016/S0304-405X(99)00057-4). *Journal of Financial Economics*.
- **Thomas Ho and Hans R. Stoll** (1981). [Optimal dealer pricing under transactions and return uncertainty](https://doi.org/10.1016/0304-405X(81)90020-9). *Journal of Financial Economics*.
- **Xavier Gabaix and Ralph S. J. Koijen** (2021). [In Search of the Origins of Financial Fluctuations: The Inelastic Markets Hypothesis](https://doi.org/10.3386/w28967).
- **Yakov Amihud** (2002). [Illiquidity and stock returns: cross-section and time-series effects](https://doi.org/10.1016/S1386-4181(01)00024-6). *Journal of Financial Markets*.
- **Yuki Sato and Kiyoshi Kanazawa** (2024). [Does the square-root price impact law belong to the strict universal scalings?: quantitative support by a complete survey of the Tokyo stock exchange market](https://doi.org/10.48550/arXiv.2411.13965).
- **Yuriy Nevmyvaka et al.** (2006). [Reinforcement learning for optimized trade execution](https://doi.org/10.1145/1143844.1143929). *Association for Computing Machinery*.
- **Zoltan Eisler et al.** (2010). [The price impact of order book events: market orders, limit orders and cancellations](https://doi.org/10.48550/arXiv.0904.0900).