Overnight Inventory: Does It Predict RTH Direction?
2008-2026 | 4,580 sessions with overnight data | ES 5-min
What is Overnight Inventory?
Overnight inventorymeasures the positioning of the Globex session relative to the prior day's Value Area. When the overnight session trades predominantly above the prior VAH, it is classified as long inventory. When it trades below the prior VAL, it is short inventory. Sessions straddling the Value Area are neutral.
The core hypothesis: if the overnight session builds excess inventory in one direction, the RTH session should correct back toward value — i.e., long inventory predicts a downward correction at the open, and short inventory predicts an upward correction.
How to use this
- VWAP classification— Overnight VWAP above VAH = long; below VAL = short. This is the primary method tested.
- Close classification— Overnight close above VAH = long; below VAL = short. An alternative method for comparison.
- Correction window— We measure whether price corrects within the first 30 minutes and across the full OTC session.
- Context matters— Inventory alone is a weak signal. Combine with IB structure, gap size, and market internals.
Inventory Distribution (VWAP Classification)
Nearly half of all sessions are classified as neutral — overnight stays within the prior Value Area
Correction Rates by Inventory Type
| Inventory | n | % of Total | First 30m Correction | OTC Direction Correct |
|---|---|---|---|---|
| Long | 1,320 | 28.8% | 58.2% | 52.1% |
| Short | 1,250 | 27.3% | 56.8% | 51.5% |
| Neutral | 2,010 | 43.9% | 50.2% | 50% |
“First 30m Correction” = price moves toward value within the first 30 minutes of RTH. “OTC Direction Correct” = full-session close in the correction direction.
Gap Fill Rates by Inventory Type
| Inventory | n | Gap Fill Rate |
|---|---|---|
| Long | 1,320 | 62.5% |
| Short | 1,250 | 60.8% |
| Neutral | 2,010 | 55.2% |
Gap fill = RTH price fills 100% of the overnight gap during the session. Long and short inventory sessions show modestly higher fill rates than neutral.
Fade Trade Backtest Variants
| Trade | n | Win% | PF | Cum PnL (pts) |
|---|---|---|---|---|
| Fade Long 4pt | 1,320 | 48.2% | 1.05 | +55.0 |
| Fade Long 6pt | 1,320 | 42.1% | 0.98 | -18.0 |
| Fade Long 8pt | 1,320 | 35.5% | 0.92 | -65.0 |
| Fade Short 4pt | 1,250 | 47.8% | 1.03 | +28.0 |
| Fade Short 6pt | 1,250 | 41.5% | 0.96 | -32.0 |
| Fade Short 8pt | 1,250 | 34.8% | 0.90 | -78.0 |
| Fade to POC | 2,570 | 52.5% | 1.08 | +145.0 |
| Fade to VA Mid | 2,570 | 55.2% | 1.12 | +210.0 |
Fixed-point fades (4/6/8 pt targets) show marginal or negative edge. Fading to POC or VA midpoint shows a modest positive expectancy — suggesting the target matters more than the entry signal.
VWAP vs Close-Based Classification
VWAP Classification
Close-Based Classification
Close-based classification produces more neutral sessions (50.2% vs 43.9%) and slightly lower correction rates across the board. VWAP classification is marginally stronger.
Yearly Correction Rate Stability
Correction rates remain in a narrow band (55–60%) across all sampled years — dashed line = 50% (coin flip)
Yearly Breakdown
| Year | n | Long Inv Corr% | Short Inv Corr% |
|---|---|---|---|
| 2008 | 169 | 60.2% | 58.5% |
| 2012 | 250 | 57.8% | 56.2% |
| 2016 | 254 | 58.5% | 57% |
| 2020 | 254 | 59.1% | 57.8% |
| 2024 | 257 | 57.2% | 55.8% |
Verdict
Weak directional signal. Not tradeable in isolation.
- Overnight inventory shows a modest correction tendency(56–58% in the first 30 minutes vs 50% baseline) when using VWAP classification. This is statistically real but economically thin.
- Fixed-point fade trades produce marginal-to-negative cumulative PnL. Only fades targeting POC or VA midpoint show a mild positive expectancy, suggesting the target (not the entry signal) is doing the heavy lifting.
- Close-based classification is weaker than VWAP-based, with more sessions falling into the uninformative neutral bucket. VWAP is the better method if you use inventory at all.
- The effect is stable across yearsbut never rises above 60% correction rate — consistently a coin flip with a slight thumb on the scale.
- Bottom line:Use overnight inventory as one input in a multi-factor framework (IB structure, gap size, market internals), not as a standalone directional predictor. The popular narrative that “long inventory above VAH means short at the open” is directionally correct but barely actionable.