orderflow-rs

A high-performance Rust pipeline for limit order book (LOB) microstructure research.

Ingests raw tick data, computes order-flow imbalance (OFI) features and 87 technical indicators, runs synthetic LOB simulations, backtests signal-based strategies, and produces ranked IC reports — all from a single binary.

Companion research: "Predictability of High-Frequency Limit Order Book Dynamics"
Full thesis: docs/paper/orderflow_research.pdf

---

Research Summary

Phase	Question	Verdict
P4 — Simulation	Does OFI_1 predict synthetic price moves?	PASS (IC 0.15–0.33)
P5 — Real data	Does OFI_1 predict real FX returns?	PARTIAL (5/10 pairs; IC 0.04–0.10)
P6 — Backtest	Can OFI signals generate net positive PnL?	FAIL (negative after 0.04 % fee + half-spread)
P7 — Tech indicators	Which classical indicators best predict 1–300 s returns?	pivot_dist IC = 0.40@1 s, power-law decay τ^−0.43

---

Features

• Ingest — streams Dukascopy LOB tick data, reconstructs bid/ask ladder, labels forward returns (1 s / 5 s / 30 s / 300 s)
• OFI — computes multi-level order-flow imbalance features (OFI_1 … OFI_5, cumulative)
• 87 technical indicators — momentum, mean-reversion, trend, volatility, microstructure, and market-impact measures
• IC engine — Spearman rank IC for every indicator × horizon combination; cross-pair averaging
• Simulation — synthetic LOB with configurable spread / queue dynamics for hypothesis testing
• Backtest — walk-forward OOS evaluation with transaction-cost accounting
• Feature flags — ingest, io (Parquet), sim are optional; core analysis compiles with zero async deps

---

Quick Start

Prerequisites

• Rust 1.75+ (rustup update stable)
• Feature data CSVs in data/ (see Data)

Build

cargo build --release

Run pipeline phases

# Ingest raw tick data → feature CSVs
cargo run --release -- ingest data/

# Compute OFI features
cargo run --release -- features data/

# Run synthetic LOB simulation (P4)
cargo run --release -- simulate

# Backtest OFI signals (P6)
cargo run --release -- backtest data/ reports/

# Technical indicator IC analysis (P7)
cargo run --release -- techanalysis data/ reports/

Reports are written to reports/ as CSV files (gitignored — run locally to reproduce).

---

Project Structure

orderflow-rs/
├── src/
│   ├── main.rs                  # CLI entry point
│   ├── commands/                # Subcommand handlers
│   │   ├── techanalysis.rs
│   │   ├── backtest.rs
│   │   └── ...
│   ├── analysis/
│   │   ├── techind.rs           # 87 indicator implementations
│   │   ├── techreport.rs        # IC computation & CSV output
│   │   ├── stats.rs             # Spearman IC, variance ratio, Hurst
│   │   ├── report.rs            # CSV loader / feature row schema
│   │   └── ...
│   ├── features/                # OFI feature extraction
│   ├── orderbook/               # LOB reconstruction
│   ├── simulator/               # Synthetic LOB
│   └── pipeline/                # Ingest orchestration
├── tests/                       # Integration & property tests
├── benches/                     # Criterion benchmarks
├── docs/
│   ├── paper/                   # LaTeX source + compiled PDF
│   └── reports/                 # Markdown phase reports
├── Cargo.toml
└── Cargo.lock

---

Data

Raw tick data is not included in this repository (files are too large and subject to Dukascopy terms of use).

To reproduce the analysis:

1. Download Dukascopy LOB tick data for the target FX pairs (EURUSD, GBPUSD, USDJPY, etc.)
2. Place raw files under data/<source>/<PAIR>/
3. Run cargo run --release -- ingest data/ to produce *_features.csv files

Pre-processed feature CSVs (187 files, ~2 GB) are available on request.

---

Reproducing Results

# P5 — OFI IC analysis
cargo run --release -- analyze data/ reports/

# P6 — Walk-forward backtest
cargo run --release -- backtest data/ reports/

# P7 — Technical indicator IC ranking
cargo run --release -- techanalysis data/ reports/

All results are deterministic given the same input CSVs. See docs/paper/orderflow_research.pdf Appendix B for full reproduction instructions.

---

Testing

cargo test                        # unit + integration tests
cargo test --features sim         # includes simulation tests
cargo bench                       # Criterion benchmarks

---

License

MIT — see LICENSE.

---

Citation

@thesis{tkoovonzko2025orderflow,
  author  = {Tkoovonzko, Evin},
  title   = {Predictability of High-Frequency Limit Order Book Dynamics},
  school  = {},
  year    = {2025},
  url     = {https://github.com/Evintkoo/orderflow-rs}
}