Hurst/VPIN Directional

A directional example strategy that combines a Hurst-exponent regime filter on dollar bars with a VPIN (Volume-synchronized Probability of Informed Trading) signal from trade aggressor flow. Entry timing is gated on the live quote stream.

Strategy overview

The strategy runs three concurrent pipelines over a single instrument:

Per trade: aggressive buy and aggressive sell volume is accumulated into the current dollar-bar bucket, using TradeTick::aggressor_side.
Per bar (bucket close): the bar's log return joins a rolling Hurst window, the bucket's signed and absolute imbalance joins a rolling VPIN window, accumulators reset, and the Hurst and VPIN signals are re-estimated.
Per quote: when the signals are warm and the strategy is flat, a trending Hurst and an informed VPIN of consistent sign trigger an IOC market order in the direction of the signed imbalance.

Exit fires from the bar pipeline on regime decay (Hurst drops below the exit threshold) and from the quote pipeline on a holding-time cap.

Signals

Hurst exponent by rescaled range over hurst_window dollar-bar log returns, regressed across the lag set in hurst_lags. Values above hurst_enter indicate a persistent (trending) regime; values below hurst_exit indicate regime decay.
VPIN defined on dollar-bar buckets. For each completed bar with nonzero aggressor-classified volume:
```
imbalance = (buy_volume - sell_volume) / (buy_volume + sell_volume)
```
VPIN is the mean of |imbalance| over the last vpin_window buckets. A signed variant retains the sign of buy - sell and carries the net informed direction.

Sampling frame

Dollar bars (VALUE aggregation) are the natural sampling frame for both signals, following Lopez de Prado (Advances in Financial Machine Learning, Chapter 2). Each bar closes after a fixed notional has traded, so the sampling frame adapts to market activity, and VPIN volume buckets coincide with the bar boundaries.

Parameters

Parameter	Type	Default	Description
`instrument_id`	`InstrumentId`	required	Instrument to subscribe to and trade.
`bar_type`	`BarType`	required	Dollar bar type (`VALUE` aggregation, `LAST` prices).
`trade_size`	`Quantity`	required	Order quantity for each entry.
`hurst_window`	`usize`	`128`	Rolling window of dollar bar log returns.
`hurst_lags`	`Vec<usize>`	`[4, 8, 16, 32]`	Lag set used in the R/S regression.
`hurst_enter`	`f64`	`0.55`	Above this, the regime is treated as trending.
`hurst_exit`	`f64`	`0.50`	Below this, open positions are flattened.
`vpin_window`	`usize`	`50`	Number of volume buckets averaged for VPIN.
`vpin_threshold`	`f64`	`0.30`	Minimum VPIN for flow to be considered informed.
`max_holding_secs`	`u64`	`3600`	Maximum seconds a position may be held.

Rust usage

rust

use nautilus_model::{
    data::BarType,
    identifiers::InstrumentId,
    types::Quantity,
};
use nautilus_trading::examples::strategies::{
    HurstVpinDirectional, HurstVpinDirectionalConfig,
};

let instrument_id = InstrumentId::from("PF_XBTUSD.KRAKEN");
let bar_type = BarType::from("PF_XBTUSD.KRAKEN-2000000-VALUE-LAST-INTERNAL");

let config = HurstVpinDirectionalConfig::new(
    instrument_id,
    bar_type,
    Quantity::from("0.01"),
);

engine.add_strategy(HurstVpinDirectional::new(config))?;

Extending this strategy

This strategy is intentionally minimal. Common extensions:

Volatility gate: overlay a realized-volatility estimator on the same bars to suppress entries during clearly chaotic sessions.
Alternative bars: swap VALUE for VALUE_IMBALANCE or VALUE_RUNS to sample directly on information arrival.
Risk budgeting: scale trade_size by current portfolio exposure rather than using a fixed quantity.
Multi-instrument: run the strategy across a basket and allocate across signals by Hurst strength.