Decoding Geopolitical Risk: A Level 3 Market Microstructure Analysis of Crude Oil at the Start of the Iran War

First published by TabbFORUM 

By Roger Ahanonu, Head of Data Science (EMEA & APAC)

Geopolitical risk in financial markets is typically understood as a macro phenomenon - stress escalates, supply is threatened, prices spike. But the order book tells a different story. Using BMLL Level 3 data, containing every insert, cancel, and execution at every price level, we show that the market's response to the 2026 Iran war was visible in microstructure signals days before it appeared in price.

We show that order flow Imbalance revealed directional selling pressure 10 days before the war onset. Order book depth collapsed 80% from pre-war baselines, with the ask side falling first, a signature of informed traders anticipating supply disruption. Bid-ask spreads widened to 7× their normal levels and persisted long after headlines normalized. Market makers didn't withdraw; they engaged in continuous repricing at unprecedented intensity, processing 11.2M inserts in a single session at peak stress. This is one of the most extreme microstructure stress episodes in crude oil futures in at least three decades, and price charts alone would have missed most of it.

Context

On 28 February 2026, joint U.S. and Israeli airstrikes struck multiple sites across Iran. Over the 12 days that followed, the region experienced sustained missile and drone exchanges, strikes on Gulf oil infrastructure, refinery disruptions, and credible Strait of Hormuz mining threats — culminating in Brent crude hitting an intraday peak of $119.50/bbl on March 9.

Price charts tell part of this story. Level 3 order book data tells the rest.

We utilize BMLL Futures L3 Orders — full order lifecycle events including every insert, cancel, and execution at every price level — to answer questions that L1 or L2 data alone cannot:

• How fast did market makers pull quotes after each headline? (Cancellation-to-Quote Ratio)
• What was the order book depth profile across all 10 levels during war onset? (L2 Depth and Heatmap)
• Did bid-ask spreads widen differently for supply shocks vs geopolitical uncertainty? (TWA Spread)
• Which exchange lead price discovery — ICE Brent or CME WTI — and how did that shift as war escalated? (Lead score)
• Was there measurable Order Flow Imbalance ahead of the price move? (OFI leading indicator)

Three events are analyzed across a graduated severity spectrum: E4 (Hormuz precursor, Feb 20), E5 (war onset, Feb 28 — primary focus), and E6 (peak market stress, Mar 9 — Brent $119.50/bbl).

1. Market Makers Pulled Before the Largest Rise

The order book doesn't just empty because people trade, it empties because market makers cancel their quotes. This distinction is precisely what L3 data makes visible. The Cancellation-to-Quote Ratio (CQR), the number of orders pulled per order placed, remained remarkably stable throughout the pre-war period at 0.94–0.96, but the absolute volume of activity told a different story.

The progression reveals three distinct phases:

• Pre-war baseline (Feb 3–19): CQR stable at 0.95, daily inserts averaging 3.0M–4.0M
• Threat escalation (Feb 20–27): CQR holds steady at 0.94–0.96, but Feb 27 shows a 2.4× surge in absolute cancellations (2.66M) relative to the previous week's average
• War onset and peak stress (Mar 1–9): CQR remains at 0.94–0.96, yet daily insert activity explodes to 11.2M on Mar 9, a 3.7× increase over the pre-war baseline, with cancellations tracking proportionally at 10.8M

The critical insight: CQR stability masks a complete regime change in market maker behavior. Market makers didn't become more defensive (CQR unchanged); they became hyper-active, continuously repricing and repositioning at unprecedented scale. The book didn't empty, it churned.

At E6 (Mar 9), when Brent touched $119.50, the market processed 11.2M inserts and 10.8M cancels; more than double the busiest pre-war session. This is not withdrawal; this is continuous repricing under extreme uncertainty.

2. Depth Collapsed Asymmetrically, Then Uniformly

Under normal conditions, the Brent order book carries substantial advertised depth across all 10 price levels. Time-Weighted Average (TWA) depth, which accounts for how long liquidity actually sits in the book, not just snapshots, reveals how market makers repositioned during the crisis.

Pre-war baseline (Feb 3–26):

• Bid-side TWA: 130–195 contracts across L1–L10
• Ask-side TWA: 135–175 contracts
• Total TWA depth: 270–360 contracts (combined bid + ask)

E4 precursor (Feb 20):

• Bid TWA: 167.6 | Ask TWA: 153.5 | Total: 321.1
• Spreads begin to widen: 0.0214 (vs. 0.017–0.022 baseline)

War onset E5 (Feb 28–Mar 1):

• Asymmetric collapse: Ask-side TWA plummets to 56.96 on Mar 1 (a 67% decline), while bid-side TWA initially rises to 295.2; market makers pulled sell liquidity while buyers aggressively defended downside
• Spread explodes to 0.0292 (36% wider than E4)

Peak stress E6 (Mar 9):

• Bid TWA: 34.27 | Ask TWA: 37.20 | Total: 71.47 — an 80% collapse from pre-war levels
• Both sides now equally depleted, indicating complete market maker withdrawal
• Spread peaks at 0.1270 — 7× the pre-war average and the widest spread observed in the entire dataset

Recovery phase (Mar 10–11):

• Depth begins rebuilding slowly: Total TWA recovers to 112.9 by Mar 11, but remains 65% below baseline
• Spread compression lags: Still at 0.0439 on Mar 11 (2× pre-war levels)

Key finding: The initial collapse was directional (ask-side first), revealing informed traders anticipating supply disruption. Only after war confirmation did the bid-side also collapse, creating symmetric illiquidity.

3. Insert Sizes Compressed Across All Levels

Average insert order size by price level shows how market makers manage adverse selection risk. Generally, for order books (although depending on the instrument) orders at the touch are smaller (managing execution risk), while orders deeper in the book are larger (committed capital away from immediate execution).

Pre-war pattern (Feb 3–19):

• L1 (best bid/ask): 1.9–2.4 contracts per insert
• L5–L10 (deep levels): 1.5–2.0 contracts — slight compression away from touch

E4 precursor (Feb 20):

• L1: 2.17 | L10: 1.94 — depth gradient intact but compressing

Pre-war stress signal (Feb 27):

• Anomaly detected: L5–L10 insert sizes surge to 2.7–4.7 contracts — market makers aggressively adding depth away from touch, anticipating volatility
• L1 remains lean at 1.86, suggesting defensive positioning at the best prices

War onset E5 (Mar 1–6):

• Complete inversion: The depth gradient disappears
• L1: 1.92 | L10: 1.21 on Mar 1 — deeper levels now smaller than the touch
• By Mar 6: L1: 1.36 | L10: 2.73 — extreme fragmentation, no consistent pattern

Peak stress E6 (Mar 9):

• Uniform compression: L1: 1.16 | L10: 1.87 — all levels thin, but the gradient has flattened
• Market makers are unwilling to commit size anywhere in the book

Key insight: The Feb 27 surge in deep-level insert sizes was a pre-emptive defensive repositioning by informed market makers. They added depth at L5–L10 to cushion against anticipated volatility, then systematically withdrew it once war began.

Below we see the ratio of average insert size deep in the book (L5–L10) to average insert size at the touch (L1–L2). In a calm market this is typically 3–6×, deep orders are naturally larger because they face lower execution probability and lower adverse selection risk.

4. Spreads Widened and Persisted

The daily time-weighted average bid-ask spread at the touch captures the execution cost premium the market charges during stress. Unlike snapshots, TWA spread accounts for how long wide spreads persist, a critical distinction during volatile periods.

Pre-war baseline (Feb 3–26):

• Median TWA spread: 0.0186 (1.86 basis points)
• Range: 0.0155–0.0270 — spreads occasionally widen intraday but quickly normalize

E4 (Feb 20):

• TWA spread: 0.0214 — a 15% widening, signaling early stress but still within normal bounds

War onset E5 (Feb 28–Mar 1):

• Feb 28 (Saturday, markets closed): N/A
• Mar 1 (Monday open): 0.0292 — a 57% jump from baseline, the largest single-session move in the dataset

Escalation phase (Mar 2–8):

• Spreads remain elevated but initially compress: 0.0315 → 0.0284 (Mar 2–4)
• Mar 6: 0.0408 — a secondary widening as conflict intensifies

Peak stress E6 (Mar 9):

• TWA spread: 0.1270 — 7× the pre-war baseline
• This is not a flash event; the spread stayed wide throughout the session, indicating sustained illiquidity

Recovery lag (Mar 10–11):

• Mar 10: 0.0881 (still 4.7× baseline)
• Mar 11: 0.0439 (still 2.4× baseline)

Key finding: Spreads widened in two distinct waves — the first at war confirmation (Mar 1), the second at peak physical disruption (Mar 9).

5. Order Flow Imbalance Led the Move

The OFI directionality ratio, the net directional component of L1 order activity, normalized for overall intensity, reveals whether informed traders are accumulating or distributing ahead of price moves. A negative OFI indicates net selling pressure at the best bid; a positive OFI indicates net buying pressure at the best offer.

Pre-war baseline (Feb 3–19):

• OFI ratio oscillates between -0.03 and +0.01 - normal two-sided flow, no persistent directional bias
• Gross OFI (absolute order flow intensity) averages 2.5M–3.2M per session

E4 precursor (Feb 20):

• OFI ratio: -0.0071 -slight negative bias, but within normal range
• Gross OFI: 2.60M - unremarkable

Pre-war signal (Feb 11, Feb 23, Feb 27):

• Feb 11: -0.0598 -strongest selling pressure of the pre-war period
• Feb 23: -0.0361 - renewed selling
• Feb 27: -0.0185 - sustained negative OFI as war onset approaches

War onset E5 (Mar 1):

• OFI ratio flips positive: +0.0274 - informed buyers enter aggressively
• Gross OFI: 1.06M - lower absolute flow, but highly directional

Peak stress E6 (Mar 9):

• OFI ratio: +0.0222 - sustained buying pressure at the offer
• Gross OFI: 2.96M -  2.8× the Mar 1 level, indicating broad participation

Recovery (Mar 10–11):

• OFI ratio: +0.0133 → +0.0057 - buying pressure dissipates but remains positive

Key insight: OFI drifted persistently negative in the 10 days before war onset, revealing informed selling pressure building before the disruption was public. The flip to positive OFI on Mar 1 coincided exactly with war confirmation - informed traders had already positioned, and the market chased them higher.

This is an L3-exclusive signal. L1 or L2 data would show a book that looked roughly normal in those pre-event sessions. Only the full order lifecycle reveals the directional intent accumulating beneath the surface.

6. Price Discovery: WTI Led First, Brent Flipped at Confirmation

The received wisdom is that Brent leads during Gulf supply shocks because it is the seaborne benchmark. The data tells a more nuanced story.

We look at the lead score. This score, calculated from the difference between two 1-minute cross-correlations (past Brent predicting current WTI vs. past WTI predicting current Brent), measures market leadership. A positive score means Brent is the price leader, transmitting information to WTI. A negative score indicates the reverse: WTI leads Brent. Averaged daily from a rolling 30-minute window, it reflects the consistency of the leadership throughout the trading day.

Pre-war baseline (Feb 3–19):

• Correlation: 0.90–0.98 — Brent and WTI are tightly coupled, as expected for two highly cointegrated crude benchmarks
• Lead score oscillates between −0.01 and +0.01 — no dominant leader under normal conditions

Threat escalation (Feb 20–27):

• Feb 20 (E4): Lead score −0.002, correlation 0.939 — no immediate leadership shift at the Hormuz threat
• Feb 27: Lead score −0.036, correlation collapses to 0.813 — the lowest co-movement of the dataset as war risk peaks. WTI takes a slight informational lead

War onset E5 (Feb 28–Mar 1):

• Feb 28 (Saturday): Markets closed
• Mar 4: Lead score −0.031, correlation 0.892 — WTI continues to lead modestly as the market processes the shock

Peak stress E6 (Mar 9):

• Mar 8: Lead score +0.129 — the strongest Brent leadership reading of the dataset. Brent decisively reclaims the informational lead as the crisis deepens
• Mar 9 (E6): Lead score +0.003, correlation 0.921 — leadership neutralises, prices re-synchronise

Recovery (Mar 10–11):

• Correlation recovers to 0.951– 0.956 — tightest co-movement since mid-February
• Lead score oscillates near zero — no persistent leader, normal regime restored

The dominant signal is correlation, not leadership. Pre-war, Brent and WTI move in lockstep (ρ ≈ 0.94). As war risk builds through late February, that coupling fractures reaching 0.81 on Feb 27. The two benchmarks temporarily decouple as participants price geographically distinct supply risks. Co-movement only fully recovers by Mar 10–11.

On leadership: the scores are small throughout, reflecting the difficulty of consistently leading a tightly arbitraged pair. The clearest signal is the Mar 8 spike (+0.129), where Brent briefly but decisively led WTI into the peak stress window consistent with informed flow shifting to the contract directly exposed to the Strait of Hormuz disruption.

Rather than only looking between WTI and Brent, we examine the lead score between the Brent/WTI spread (traded on IFEU) and the Brent outright. The Brent/WTI spread (WBS) is a natural venue for relative-value expression. A participant with a view on the differential supply impact to Brent vs WTI would trade the spread rather than either outright. Throughout the pre-war and threat phase (Feb 3–27), the Brent outright led the spread consistently, with the lead score persistently negative (−0.02 to −0.09). Even E4 (Hormuz threat, Feb 20) produced no shift.

The regime break came at peak stress. On Mar 8, the WBS lead score spiked to +0.346, by far the largest reading in the dataset, followed by +0.100 at E6 (Mar 9). For two sessions, the spread market was transmitting geopolitical pricing information before the outright adjustment. By Mar 10 the outright had reasserted leadership and the relationship normalised.

Conclusion

The Iran war of February - March 2026 produced one of the  most extreme crude oil microstructure stress episodes in at least three decades. This analysis, built on BMLL Level 3 order data, reveals a coherent and quantifiable story that price charts alone cannot tell.

What we found:

1. Cancellation activity surged 3.7×, but CQR remained stable at 0.94–0.96. Market makers didn't withdraw - they engaged in continuous repricing at unprecedented intensity. At E6 (Mar 9), the market processed 11.2M inserts and 10.8M cancels, more than triple the pre-war baseline.
2. Order book depth collapsed 80% from baseline, asymmetrically at first. Ask-side TWA depth fell 67% ahead of war onset (Feb 28–Mar 1), while bid-side depth initially surged before collapsing uniformly at E6 (Mar 9). Recovery was slow: depth remained 65% below baseline a week later.
3. Bid-ask spreads widened 7× at peak stress and persisted. TWA spread jumped from 0.0186 (baseline) to 0.0292 at war onset (Mar 1), then spiked to 0.1270 at E6 (Mar 9) - the widest spread in the entire dataset. Spreads remained 2.4× baseline even two days into recovery.
4. Insert order size inverted the normal depth gradient. Pre-war, deeper levels (L5-L10) carried larger orders (1.5-2.0 contracts). On Feb 27, deep-level insert sizes surged to 2.7- 4.7 contracts - a defensive repositioning signal. At war onset, the gradient disappeared entirely, with all levels compressing to 1.2–1.4 contracts by Mar 9.
5. OFI showed directional selling pressure 10 days before war onset. OFI ratio drifted persistently negative (Feb 11: -0.0598, Feb 23: -0.0361, Feb 27: -0.0185), revealing informed traders distributing ahead of the event. The flip to positive OFI (+0.0274) on Mar 1 coincided exactly with war confirmation.
6. Price discovery between Brent and WTI showed no consistent leader until peak stress. Throughout the pre-war and threat phase (Feb 3–27), lead scores oscillated near zero - neither benchmark dominated. Even at E4 (Hormuz threat), the lead score was only −0.002. The clearest signal came at peak stress: Mar 8 produced a lead score of +0.129, the highest Brent-leads reading of the dataset, as informed flow shifted to the contract directly exposed to the physical disruption. By E6 (Mar 9) leadership had already neutralised, and correlation recovered to 0.95+ by Mar 10–11.

The Broader Lesson

Geopolitical risk in energy markets is not just a price phenomenon — it is a microstructure phenomenon. The order book, not the price chart, is the primary leading indicator:

• Depth collapse preceded price spikes by days
• OFI signaled directional intent before public confirmation
• Spread widening persisted long after headlines normalized
• Price discovery leadership shifted the session before peak price

BMLL's L3 data infrastructure makes this level of analysis tractable for the first time at scale. For market participants navigating geopolitical shocks, the order book is the earliest and most reliable signal — and L3 data is the only way to see it in detail.

Markets Analysed: Brent Crude (ICE Futures Europe / IFEU) & WTI Crude (CME Group / XNYM)
Analysis window: 2026-02-01 to 2026-03-11
Data source: BMLL Technologies — Futures-Reference, L2/L3