Net taker flow over the trailing four hours

Net taker flow over the trailing four hours as a share of its traded notional, the same signed lean as taker_net_ratio_1h read at the horizon a move is judged on.

Observed: trailing_4h
Unit: ratio
Valid range: -1 to 1

In plain English

Measures the net buyer-versus-seller initiation of traded notional across the trailing four hours.

How to read it

Sum taker buy and sell USD across (t - 4h, t], then divide their difference by their sum. +0.2 means 60% buy-initiated notional, -0.2 means 60% sell-initiated, and zero means balanced. This is a volume-weighted four-hour ratio, not the simple average of four hourly ratios. It is absent before the full observation window is covered or when the window has no traded notional.

Constructed example · not historical data

The four-hour total weights traded notional

First 3 hours: $30k buys + $10k sellsNet +0.5
Last hour: $10k buys + $30k sellsNet -0.5
Full 4 hours: $40k buys + $40k sellsNet 0

Filled portions show buy share within each row. Despite lasting three times longer, the first period has the same notional as the last hour here. They cancel in the four-hour total; time alone supplies no extra weight.

Why a researcher might use it

Compare the broader flow window with the latest hour. Different signs can show that the latest hour's lean differs from the four-hour total; the windows overlap.

What it does not prove

A balanced four-hour total can hide strongly opposed shorter bursts. It does not measure new capital, position openings or the persistence of future flow, and overlapping windows are not independent evidence.

Use in Research

Technical details

Field ID

feature.taker_net_ratio_4h

Valid query operators

at least, at most, between

Window operators: lowest, highest, average, latest or change in, over a trailing window.

Editable query preset

This preset is a starting point, not a recommendation. Edit it before running the search.

{
  "schema_version": "research_query.v2",
  "normalization_version": "archive_normalization.v1",
  "feature_version": "feature_defs.v1",
  "target": "record_occurrences",
  "where": {
    "all": [
      [
        "feature.taker_net_ratio_4h",
        "gte",
        0.2
      ],
      [
        "times.anchor_time",
        "gte",
        "2025-07-15T00:00:00.000Z"
      ]
    ]
  },
  "sort": [
    "times.anchor_time",
    "desc"
  ],
  "page": {
    "limit": 5,
    "cursor": null
  }
}

The search reports how often the record held this condition and what followed. It does not decide whether an idea works.