Backtest Configuration

Configure backtest parameters, latency models, fees, risk limits, and execution simulation

The backtesting engine uses BacktestConfig to control all aspects of simulation realism, from network latency to fill probabilities.

BacktestConfig Structure

The main configuration is defined in nano-backtest/src/config.rs:11-29:

pub struct BacktestConfig {
    pub initial_capital: f64,
    pub latency: LatencyConfig,
    pub fees: FeeConfig,
    pub risk: RiskConfig,
    pub execution: ExecutionConfig,
    pub output: OutputConfig,
}

Initial Capital

Starting capital in USD for the backtest.

initial_capital: 1_000_000.0  // $1M default

Latency Configuration

Realistic latency simulation is critical for HFT backtesting. The LatencyConfig struct (config.rs:104-132) models network and exchange delays:

pub struct LatencyConfig {
    pub order_latency_ns: u64,           // Order submission latency
    pub market_data_latency_ns: u64,     // Market data reception latency
    pub ack_latency_ns: u64,             // Order acknowledgment latency
    pub jitter_ns: u64,                  // Latency jitter (std dev)
    pub use_random_jitter: bool,         // Enable random jitter
}

Latency Components

Component	Description	Typical Range
order_latency_ns	Time from order submission to exchange receipt	5,000 - 200,000 ns
market_data_latency_ns	Time from exchange update to strategy receipt	2,500 - 100,000 ns
ack_latency_ns	Time from exchange ack to strategy receipt	5,000 - 250,000 ns
jitter_ns	Standard deviation of latency variance	1,000 - 50,000 ns

Jitter Models

The latency simulator supports multiple jitter models (latency.rs:26-52):

• None: Deterministic latency (no variance)
• Uniform: Random jitter in range [-max_jitter, +max_jitter]
• Normal: Gaussian distribution with configurable std dev
• LogNormal: Realistic network latency distribution
• Empirical: Based on historical percentile data [p50, p75, p90, p95, p99]

Preset Latency Profiles

Aurora (CME Primary Colo) - latency.rs:215-220:

ColoLatencyModel::aurora()  // 5μs to exchange, 2μs processing

Aggressive HFT - config.rs:47-56:

order_latency_ns = 50_000       # 50 microseconds
market_data_latency_ns = 10_000 # 10 microseconds
ack_latency_ns = 60_000         # 60 microseconds
jitter_ns = 5_000               # 5 microseconds jitter
use_random_jitter = true

Conservative - config.rs:76-82:

order_latency_ns = 200_000      # 200 microseconds
market_data_latency_ns = 50_000 # 50 microseconds
ack_latency_ns = 250_000        # 250 microseconds
jitter_ns = 50_000              # 50 microseconds jitter
use_random_jitter = true

Fee Configuration

The FeeConfig struct (config.rs:134-172) models exchange and clearing fees:

pub struct FeeConfig {
    pub maker_fee: f64,      // Fee per contract for adding liquidity
    pub taker_fee: f64,      // Fee per contract for taking liquidity
    pub exchange_fee: f64,   // Exchange fee per contract
    pub clearing_fee: f64,   // Clearing house fee per contract
}

CME Default Fees

FeeConfig::default()  // Uses CME futures fees:
// maker_fee: CME_MAKER_FEE
// taker_fee: CME_TAKER_FEE
// exchange_fee: CME_EXCHANGE_FEE
// clearing_fee: CME_CLEARING_FEE

Fee Calculation

Total fee per trade:

fee = (maker_or_taker_fee + exchange_fee + clearing_fee) * quantity

Example for a 10-contract maker trade:

fees.calculate_fee(10, true)  // is_maker = true

Risk Configuration

The RiskConfig struct (config.rs:174-207) defines position limits and risk controls:

pub struct RiskConfig {
    pub max_position: i64,          // Maximum net position (contracts)
    pub max_order_size: u32,        // Maximum single order size
    pub max_drawdown_pct: f64,      // Kill switch drawdown threshold
    pub max_daily_loss: f64,        // Maximum daily loss in dollars
    pub max_open_orders: usize,     // Maximum concurrent open orders
    pub enable_kill_switch: bool,   // Enable automatic shutdown
}

Default Risk Limits

max_position = 100          # DEFAULT_MAX_INVENTORY
max_order_size = 20         # contracts
max_drawdown_pct = 0.05     # 5% drawdown limit
max_daily_loss = 100_000.0  # $100k daily loss limit
max_open_orders = 20        # concurrent orders
enable_kill_switch = true   # auto-stop on breach

Risk Presets

Aggressive HFT:

max_position = 50
max_order_size = 10
max_drawdown_pct = 0.04  # 4%
max_daily_loss = 50_000.0

Conservative:

max_position = 20
max_order_size = 5
max_drawdown_pct = 0.03  # 3%
max_daily_loss = 30_000.0

Execution Configuration

The ExecutionConfig struct (config.rs:209-242) controls fill simulation realism:

pub struct ExecutionConfig {
    pub track_queue_position: bool,       // Track position in order queue
    pub fill_probability_decay: f64,      // Fill prob decay with queue depth
    pub partial_fill_probability: f64,    // Probability of partial fills
    pub min_partial_fill: u32,            // Minimum partial fill size
    pub simulate_adverse_selection: bool, // Model adverse selection
    pub adverse_selection_prob: f64,      // Adverse selection probability
}

Fill Simulation

Queue Position Tracking (execution.rs:63-68):

• Tracks order position in the limit order book queue
• Fill probability decays exponentially: decay^queue_position
• Default decay: 0.9 (90% probability at front, 81% at position 2, etc.)

Partial Fills (execution.rs:185-198):

• Simulates realistic partial execution
• Default 20% probability of partial fill
• Fill size randomly chosen between min_partial_fill and remaining quantity

Adverse Selection (execution.rs:171-177):

• Models the tendency to get filled when market moves against you
• Default 10% probability that fill doesn’t occur due to adverse selection
• Critical for realistic maker strategy backtesting

Default Execution Settings

track_queue_position = true
fill_probability_decay = 0.9
partial_fill_probability = 0.2
min_partial_fill = 1
simulate_adverse_selection = true
adverse_selection_prob = 0.1

Conservative Preset (config.rs:93-97):

fill_probability_decay = 0.7  # More conservative fills
partial_fill_probability = 0.4 # More partial fills

Output Configuration

The OutputConfig struct (config.rs:244-273) controls logging and data recording:

pub struct OutputConfig {
    pub verbosity: u8,               // Log level (0-3)
    pub record_tick_pnl: bool,       // Record P&L at every tick
    pub record_fills: bool,          // Record all fill events
    pub record_orders: bool,         // Record all order events
    pub snapshot_interval: usize,    // Metrics snapshot frequency
}

Default Output Settings

verbosity = 1              # Basic logging
record_tick_pnl = false    # Disabled for performance
record_fills = true        # Keep fill history
record_orders = true       # Keep order history
snapshot_interval = 10000  # Snapshot every 10k events

Enabling record_tick_pnl = true can significantly impact performance and memory usage for high-frequency strategies.

Example Configuration

TOML Format

Create a backtest_config.toml file:

initial_capital = 1_000_000.0

[latency]
order_latency_ns = 100_000
market_data_latency_ns = 50_000
ack_latency_ns = 120_000
jitter_ns = 10_000
use_random_jitter = true

[fees]
maker_fee = 0.25
taker_fee = 0.85
exchange_fee = 0.10
clearing_fee = 0.02

[risk]
max_position = 50
max_order_size = 10
max_drawdown_pct = 0.05
max_daily_loss = 100_000.0
max_open_orders = 20
enable_kill_switch = true

[execution]
track_queue_position = true
fill_probability_decay = 0.9
partial_fill_probability = 0.2
min_partial_fill = 1
simulate_adverse_selection = true
adverse_selection_prob = 0.1

[output]
verbosity = 1
record_tick_pnl = false
record_fills = true
record_orders = true
snapshot_interval = 10000

Rust API

use nano_backtest::BacktestConfig;

// Use default configuration
let config = BacktestConfig::default();

// Use preset
let config = BacktestConfig::aggressive_hft();
let config = BacktestConfig::conservative();

// Custom configuration
let config = BacktestConfig {
    initial_capital: 500_000.0,
    latency: LatencyConfig {
        order_latency_ns: 75_000,
        market_data_latency_ns: 25_000,
        ack_latency_ns: 100_000,
        jitter_ns: 15_000,
        use_random_jitter: true,
    },
    fees: FeeConfig::default(),
    risk: RiskConfig {
        max_position: 30,
        max_drawdown_pct: 0.03,
        ..Default::default()
    },
    execution: ExecutionConfig::default(),
    output: OutputConfig::default(),
};

Configuration Presets

Two built-in presets are available:

Aggressive HFT (config.rs:45-69)

Optimized for sub-millisecond strategies:

• Ultra-low latency (50μs orders)
• Tight risk limits (50 max position)
• Realistic fill simulation

let config = BacktestConfig::aggressive_hft();

Conservative (config.rs:71-100)

Optimized for strategy validation:

• Higher latency (200μs orders)
• Buffered fees (1.2x CME rates)
• Conservative fill model (70% decay)
• Tighter drawdown limit (3%)

let config = BacktestConfig::conservative();

Related

• Running Backtests - Execute backtests with your configuration
• Performance Metrics - Interpret backtest results
• Risk Configuration - Risk control implementation details