TradeSignAutocorrelation

The lag-1 autocorrelation of the trade-aggressor side — how strongly signed order flow persists, a footprint of order-splitting and informed execution.

Quick reference

Field	Value
Family	Microstructure
Input type	Trade (price / size / side)
Output type	f64
Output range	[−1, +1]
Default parameters	(period = 20) (Python)
Warmup period	period
Interpretation	> 0 flow persists; < 0 bid-ask bounce; ~0 balanced.

Formula

s_t = +1 if buy, −1 if sell
ρ1  = mean over the window of ( s_t · s_{t−1} )      ∈ [−1, +1]

Trade signs are strongly positively autocorrelated in real markets: large parent orders are sliced into many same-side child trades, so a buy is usually followed by a buy. A high reading flags persistent directional pressure (informed / algorithmic execution); near zero is balanced flow; negative is alternating (bid-ask bounce). Source: crates/wickra-core/src/indicators/trade_sign_autocorrelation.rs.

Parameters

Name	Type	Default	Valid range	Source	Description
period	usize	20 (Python)	>= 2	trade_sign_autocorrelation.rs:55	Rolling window of trades. < 2 errors with Error::InvalidPeriod.

The period getter returns the window; value returns the current output if ready.

Inputs / Outputs

From crates/wickra-core/src/indicators/trade_sign_autocorrelation.rs:

use wickra::{Indicator, TradeSignAutocorrelation};
use wickra::Trade;
// TradeSignAutocorrelation: Input = Trade, Output = f64
const _: fn(&mut TradeSignAutocorrelation, Trade) -> Option<f64> =
    <TradeSignAutocorrelation as Indicator>::update;

A Trade in, an Option<f64> out. The Python binding takes update(price, size, is_buy) and batch(price[], size[], is_buy[]) (three arrays); Node mirrors update(price, size, isBuy) / batch(...). Only the side is used here.

Warmup

warmup_period() == period. The first value lands once the window holds period trade signs (first_emission_at_warmup_period pins this).

Edge cases

• Persistent flow → +1. All same-side trades give +1 (persistent_flow_is_one pins this).
• Alternating flow → −1. Strictly alternating signs give −1 (alternating_flow_is_minus_one pins this).
• Bounded. The reading stays within [−1, +1] (output_in_range pins this).
• Reset. t.reset() clears the sign window and the last value (reset_clears_state).

Examples

Rust

use wickra::{BatchExt, Indicator, TradeSignAutocorrelation};
use wickra::{Side, Trade};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut t = TradeSignAutocorrelation::new(10)?;
    let trades: Vec<Trade> = (0..20).map(|_| Trade::new(100.0, 1.0, Side::Buy, 0).unwrap()).collect();
    println!("{:?}", t.batch(&trades).last().unwrap()); // Some(1.0)
    Ok(())
}

Output:

Some(1.0)

Python

import numpy as np
import wickra as ta

t = ta.TradeSignAutocorrelation(20)
price = np.full(40, 100.0); size = np.ones(40)
is_buy = np.array([i % 2 == 0 for i in range(40)])
print(t.batch(price, size, is_buy)[-1])  # -1.0 (alternating)

Node

const ta = require('wickra');
const t = new ta.TradeSignAutocorrelation(20);
console.log('warmupPeriod:', t.warmupPeriod()); // 20

Streaming

use wickra::{Indicator, TradeSignAutocorrelation};
use wickra::Trade;

let mut t = TradeSignAutocorrelation::new(20).unwrap();
let trade_feed: Vec<Trade> = Vec::new(); // your live stream
for trade in trade_feed {
    if let Some(rho) = t.update(trade) {
        // rho high -> persistent directional pressure
    }
}

Streaming update and batch are equivalent tick-for-tick (batch_equals_streaming pins this).

Interpretation

1. Informed-flow gauge. A persistently high autocorrelation suggests a large directional order is working — fade with caution, follow with confirmation.
2. Liquidity regime. Near-zero flow with low autocorrelation is healthy two-sided liquidity.
3. Microstructure noise. Strongly negative values indicate bid-ask bounce dominating — common in quiet, tick-sized markets.

Common pitfalls

• Sign classification. Quality depends on correct aggressor-side tagging (Lee-Ready or exchange-provided); mis-signed trades blur the reading.
• Window vs. event time. This is in trade (event) time, not clock time; bursty periods fill the window faster.
• Not a price predictor alone. High persistence says flow is one-sided, not that price must continue.

References

Bouchaud, J.-P., et al. (2004), "Fluctuations and response in financial markets" (long-range autocorrelation of trade signs); Lillo, F., & Farmer, J. D. (2004).

See also

• Indicator-OrderFlowImbalance — signed flow imbalance.
• Indicator-TradeImbalance — buy/sell volume imbalance.
• Indicator-Vpin — volume-synchronised informed-trading probability.
• Indicators-Overview — the full taxonomy.