microstructure-sim

High-Performance Limit Order Book + Market Microstructure Simulator in Rust

A full market microstructure research platform — not just another matching engine.

Architecture

                    ┌─────────────────────────────────────┐
                    │           microstructure-sim         │
                    │          (binary / CLI entry)        │
                    └────────────────┬────────────────────┘
                                     │
         ┌───────────┬───────────────┼───────────────┬──────────┐
         │           │               │               │          │
    ┌────▼────┐ ┌────▼─────┐  ┌─────▼──────┐  ┌────▼────┐ ┌───▼────┐
    │orderbook│ │ market-  │  │  simulator  │  │exchange │ │  stat  │
    │  -core  │ │  maker   │  │             │  │  -feed  │ │  -arb  │
    └─────────┘ └──────────┘  └─────────────┘  └─────────┘ └────────┘
         │           │               │               │          │
    ┌────▼───────────▼───────────────▼───────────────▼──────────▼────┐
    │                         metrics-engine                         │
    └────────────────────────────────────────────────────────────────┘

Modules

Crate	Description
orderbook-core	Lock-free limit order book with price-time priority, SIMD-optimized hot paths, cache-line-aligned structures
market-maker	Avellaneda-Stoikov optimal market making with inventory risk management and quote lifecycle
simulator	Latency-aware discrete-event simulation, queue position estimation, deterministic replay
exchange-feed	Live L2/L3 order book reconstruction from Binance and Coinbase WebSocket feeds
stat-arb	Cointegration analysis (ADF, Engle-Granger), Kalman filter hedge ratios, pairs trading signals
metrics-engine	Sharpe/Sortino/drawdown, fill rate, adverse selection markouts, PnL attribution, terminal dashboard

Build

cargo build --release

Usage

Backtest Mode

cargo run --release -- backtest \
  --symbol BTC-USD \
  --strategy market-maker \
  --gamma 0.1 \
  --kappa 1.5 \
  --seed 42

Live Feed (Read-Only)

cargo run --release -- live-feed \
  --exchange binance \
  --symbol btcusdt \
  --depth 20

Paper Trading

cargo run --release -- paper \
  --exchange binance \
  --symbol btcusdt \
  --strategy market-maker

Parameter Tuning

Avellaneda-Stoikov (Market Making)

Parameter	Description	Typical Range	Effect
gamma	Risk aversion	0.01 - 1.0	Higher = wider spreads, lower inventory risk
kappa	Order arrival decay	1.0 - 100.0	Calibrate from fill data; higher = steeper fill decay
sigma	Volatility	Auto-estimated	EWMA of realized variance
T	Session duration	3600 - 86400s	Crypto: 86400 (24h), TradFi: 23400 (6.5h)

Tuning strategy:

1. Start with gamma=0.1 (moderate risk aversion)
2. Calibrate kappa from 1 week of fill data using the built-in MLE calibrator
3. Adjust gamma based on max drawdown tolerance
4. Monitor fill rate — if < 10%, reduce gamma; if > 50%, increase gamma

Statistical Arbitrage

Parameter	Description	Typical Range
R (obs_variance)	Observation noise	0.01 - 10.0
Q (state_variance)	State transition noise	1e-6 - 1e-3
entry_z	Entry z-score threshold	1.5 - 3.0
exit_z	Exit z-score threshold	0.0 - 1.0
stop_loss_z	Stop-loss z-score	3.0 - 5.0

Tuning strategy:

1. Higher Q/R ratio = faster adaptation to regime changes, more noise
2. Start with entry_z=2.0, exit_z=0.5
3. Verify cointegration half-life is 5-200 periods
4. Use Kelly fraction cap of 0.25 (quarter-Kelly for safety)

Key Design Decisions

• Fixed-point arithmetic: All prices/quantities use i64/u64 with 10^8 decimal shift — eliminates float non-determinism in the matching engine
• Lock-free data structures: crossbeam_skiplist::SkipMap for price levels, DashMap for order lookup, crossbeam-epoch for memory reclamation
• Cache-line alignment: #[repr(align(64))] on PriceLevel prevents false sharing
• Parametric queue model: Queue position estimation using p(x) = x^(1+k) cancellation bias — far more realistic than naive models
• Latency modeling: Separate feed/order/cancel latency with Gaussian, log-normal, or empirical distributions

Performance Targets

Operation	Target Latency
Order insert	< 500ns
Order cancel	< 300ns
Match (per level)	< 200ns
VWAP (20 levels)	< 100ns
Depth snapshot	< 1us

References

1. Avellaneda, M. & Stoikov, S. (2008). "High-frequency trading in a limit order book." Quantitative Finance, 8(3), 217-224.
2. Gueant, O., Lehalle, C.-A., & Fernandez-Tapia, J. (2012). "Dealing with the Inventory Risk: A Solution to the Market Making Problem." Mathematics and Financial Economics.
3. Cont, R., Stoikov, S., & Talreja, R. (2010). "A Stochastic Model for Order Book Dynamics." Operations Research.
4. Hamilton, J.D. (1994). Time Series Analysis. Princeton University Press. (Cointegration, ADF test)
5. Harvey, A.C. (1990). Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge University Press.

License

MIT