Order Book Imbalance Ratios: Quantitative Trading Thresholds

Order book imbalance measures how lopsided resting liquidity is between bids and offers. Quantified as a ratio, it produces mechanical triggers less prone to interpretation than discretionary ladder reading.

The ratio formula

For a given depth window (typically top 5 or 10 levels):

Imbalance = Sum(bid size) / Sum(offer size)

• Ratio > 1.0: bid-heavy (bullish bias)
• Ratio < 1.0: offer-heavy (bearish bias)
• Many platforms invert this; pick one convention and stick to it.

Thresholds that matter

• 3:1 imbalance sustained for 60+ seconds: a directional bias. Combine with price action — enter long on the first pullback if bid-heavy, short if offer-heavy.
• 5:1 imbalance sustained 90+ seconds: strong bias. Higher-conviction entry; demand 2:1 R:R.
• 10:1+ imbalance: often a spoof designed to trick you. Demand tape confirmation (real fills) before acting. If the heavy side pulls as price approaches, it was a spoof — fade.

Stacked imbalance

A stacked imbalance is three or more consecutive price levels on the same side each showing a 3:1+ imbalance. A stacked bid imbalance (3 price levels, each 3:1 bid-heavy) is a wall of buying interest. These act as short-term support; price often bounces off them. Trade the bounce with a stop beyond the wall.

Mechanical entry rules

Long setup

1. Top-5 bid/offer ratio ≥ 3:1 sustained 60 seconds.
2. Price pulls back to the best bid.
3. Tape shows offer-lifting (buyers aggressive).
4. Enter long at the bid. Stop: 2 ticks below the stacked imbalance. Target: next overhead HVN or 2× risk.

Short setup (mirror)

1. Top-5 offer/bid ratio ≥ 3:1 sustained 60 seconds.
2. Price rallies to the best offer.
3. Tape shows bid-hitting (sellers aggressive).
4. Enter short at the offer. Stop: 2 ticks above the stacked imbalance.

False signals and filters

• Spoof filter: require that the heavy side actually fills — at least 50% of the heavy-side size must execute within 2 minutes. If it pulls, discard the signal.
• Higher-timeframe filter: only take imbalances aligned with the 15-min trend. Counter-trend 3:1 imbalances fail more than they work.
• News filter: discard imbalances within 5 minutes of a scheduled release. Spreads and spoofing spike around news.

Imbalance ratios remove guesswork from the ladder. Demand the threshold, the duration, and tape confirmation — then execute mechanically.