Definition
#Measures the rarity of the current net taker imbalance in the active rolling window relative to its historical mean and standard deviation. Positive values indicate unusually strong buy-side imbalance, while negative values indicate unusually strong sell-side imbalance.
Formula & calculation
#(Live Mean Net Taker Imbalance - Historical Mean Net Taker Imbalance) / Historical StdDev Net Taker ImbalanceUnits & range
Z (standard deviations). Can be positive or negative.
Interpretation
#Some perpetual contracts are structurally buy-biased: their raw net taker imbalance is chronically positive. Raw NTI on those instruments produces false buy signals constantly. The z-score normalizes against the instrument's own historical imbalance distribution, so a reading of +2.5 means the current buy-side dominance is exceptional for that specific market, not just meeting a universal threshold. This is the preferred directional filter for cross-symbol scanners.
Practical usage
#Use instead of raw net taker imbalance in cross-symbol scanner rules. A buy scan using raw NTI > 20% will fire constantly on structurally buy-biased contracts. The same scan using taker imbalance z-score > 2.0 fires only when the imbalance is genuinely anomalous relative to that instrument's norm. Combine with buy volume z-score for a two-layer confirmation: unusual volume AND unusual directional skew.
Common mistakes
#Frequent interpretation traps and misuse patterns to avoid when applying this metric.
- Using raw net taker imbalance for cross-symbol comparison: structural bias makes universal thresholds unreliable across instruments.
- Checking only the magnitude and ignoring the sign. A z-score of 2.5 is bullish; a z-score of -2.5 is bearish. They require different responses.
Timeframe note
#This metric applies to rolling windows such as 5m, 15m, and 60m. The underlying definition stays the same; what changes is the time horizon used to measure it. Shorter windows react faster, while longer windows smooth noise and emphasize broader structure.
5m
Faster response to fresh changes in activity and short-horizon structure.
15m
Balanced view between responsiveness and persistence.
60m
Broader context that is slower but more stable.