The Asia/EU/US AVWAP carry is mostly random.
The defense of Asia AVWAP during US is not.

01 Universe & method

The dataset covers 467 ES futures trading days between June 11, 2024 and May 5, 2026, sampled at the standard 5-minute interval over the full Extended Trading Hours session. Sessions are bucketed in Rome (CET/CEST) for operational clarity: Asia 00:00–09:00, EU 09:00–14:00, US 15:30–22:00.

For each session a cumulative Anchored VWAP is computed from the session open. The Asia AVWAP keeps accumulating beyond the Asia window and remains active through EU and US — so the same live-engine convention used by the TLADe terminal.

A session is classified CONFIRM when ≥80% of its bars (after the first hour) close on the side expected from the anchor session's direction, DISCONFIRM when ≥80% close on the opposite side, and left unclassified ("mixed") otherwise. This is a stricter criterion than the legacy AVWAP-cross sign-flip and produces ~37% mixed days that fall outside the classification.

02 The Asia × EU × US matrix

This is the natural first question: does the chain Asia → EU → US transmit information? If yes, certain combinations of Asia direction and EU outcome should reliably predict the US outcome.

Read this carefully: no cell has p < 0.05. The aggregate base rate of P(US CONFIRM EU) is 50.8% — a coin flip. Conditioning on EU outcome moves the needle by no more than ±4 percentage points.

The interesting near-significant cell is Asia SHORT × EU DISCONFIRM at 64.3% (p=0.09) — the so-called "Asia-Reversal Long" setup that was widely cited at 79% under the legacy criterion. We dissect it in §5.

03 The three robust setups

When we condition on asia_defense — the categorical "what happened when US price touched the Asia AVWAP level" — three setups emerge that survive the Wilson 95% lower bound floor. They all describe the same underlying physics: Asia AVWAP as the day's pivot, with US price action validating or invalidating it.

Setup A · Continuation on HOLD

ASIA LONG EU CONFIRM HOLD US CONFIRM

Asia ramps long. EU confirms the long bias. During US, price retests Asia AVWAP from above and closes on the Asia side (= the level held). The day continues long. US confirms EU's continuation of Asia.

sample

n = 30

hit rate

76.7%

95% CI

59.1–88.2

p-value

0.0062

Example: 2025-04-07 — Asia LONG → EU CONFIRM → HOLD → US CONFIRM. Asia AVWAP (cyan dashed) is tested from above during US and the close stays on the long side.

Setup B · Long-day reversal on BREAK

ASIA LONG EU CONFIRM BREAK US DISCONFIRM

Asia and EU both align long. During US, price drops back through Asia AVWAP and closes on the opposite side. This is not a temporary dip — it is the day's complete reversal. US disconfirms EU.

sample

n = 27

hit rate

81.5%

95% CI

63.3–91.8

p-value

0.0021

Example: 2025-04-08 — Asia LONG → EU CONFIRM → BREAK → US DISCONFIRM. The Asia AVWAP support fails decisively during US.

Setup C · Short-day reversal on BREAK

ASIA SHORT EU CONFIRM BREAK US DISCONFIRM

Mirror of Setup B. Asia and EU align short. During US, price rallies back through Asia AVWAP from below and closes on the opposite side. The full short-to-long reversal is the modal outcome.

sample

n = 28

hit rate

78.6%

95% CI

60.5–89.8

p-value

0.0046

Example: 2026-03-03 — Asia SHORT → EU CONFIRM → BREAK → US DISCONFIRM. Symmetric break through resistance, then full reversal.

Direction-coherent but under-sampled

Three more setups point in the right direction but have n < 15 and are not yet claimable as statistical findings. The pattern is consistent — they're additional evidence of the same underlying physics, awaiting confirmation from a larger dataset:

Snapshot example of the "never tested = pure trend" pattern:

2026-02-06 — Asia LONG → EU CONFIRM → UNTESTED → US CONFIRM. Price stays well above Asia AVWAP all day, never giving the level a chance to act as a pivot.

04 When US disconfirms EU, Asia almost always breaks

Conditional question: given a day where EU confirms Asia's bias but then US reverses against EU, how does the Asia AVWAP behave? Does the level attract price back to it (magnet), or is it crossed on the way to a complete regime change (transit)?

Universe: 58 days (Asia LONG/SHORT + EU CONFIRM + US DISCONFIRM EU).

The numbers are nearly perfectly symmetric across long and short and tell a sharp story:

• 97% of the time price actually reaches the Asia AVWAP level (only 1 untested day in each direction).
• 73–79% of those tests result in a full break — the close ends on the opposite side of Asia.
• Only 18–23% of tests turn into rejections where the level holds as a magnet.

05 Busted myths

Myth #1 · "Asia-Reversal Long" at 79%

ASIA SHORT EU DISCONFIRM EU.DIR LONG US CONFIRM

The popular setup: Asia is short, EU reverses against it and turns long, US continues long. Reported around 79% confirm rate under the legacy AVWAP-cross criterion. Under the price-consolidation criterion, the picture changes:

legacy rate

79.1%

re-measured

65.8%

95% lower bound

49.9%

p-value

0.0744

The legacy criterion counted coincidental AVWAP crossings as confirmations even when price didn't actually consolidate on the new side. Under the stricter price-consolidation criterion the rate drops by 13 percentage points and the 95% lower bound sinks below 50%. Directionally positive but not statistically robust in isolation. The cell becomes significant only when intersected with the defense dimension from §3.

2026-03-09 — Asia SHORT → EU DISCONFIRM → EU.dir LONG → US CONFIRM. The setup does occur, but at base rates that overlap a coin flip's 95% interval.

06 Structural up-bias

P(EU CONFIRM | Asia direction):

The asymmetry is sharp: Asia LONG continues into EU 71.6% of the time (highly significant), while Asia SHORT is indistinguishable from random. This is the signature of a structurally bullish market over the 2024–2026 period — overnight long biases tend to roll through European hours; overnight short biases get shaken off by European participants nearly half the time.

07 Limits of the study

1. Bull-skewed dataset. 2024–2026 is structurally bullish on ES. The LONG/SHORT asymmetry described in §6 may be a regime artifact. To separate regime-effect from intrinsic edge, the study should be replicated on 2022 (bear market) or on a longer multi-cycle window.
2. Defense computed only on EU-CONFIRM days. The current CSV does not populate asia_defense on EU-DISCONFIRM days. This blocks the symmetric question "when EU disconfirms Asia, does price still come back to the level?" Lightweight refactor planned.
3. Small samples on the mirror cells. Asia SHORT × HOLD (n=14) and both UNTESTED cells (n=8 and n=7) are directionally coherent with the unified rule but cannot be claimed as statistical findings. To be revisited with a longer dataset.
4. ~37% of days are "mixed." The price-consolidation criterion is strict by design — 172 days out of 467 don't consolidate enough on either side. All statistics in the article are computed on the 295 classifiable days.
5. Directional resolution rate, not P&L. The percentages reported are "% of days where the pattern resolved in the predicted direction by US close." Whether such a directional edge would translate into a profitable trading rule depends on assumptions (commissions, slippage, holding window, position sizing) that are outside the scope of this descriptive study.

08 Methodology & reproducibility

Session windows (Rome)

Asia 00:00–09:00, EU 09:00–14:00, US 15:30–22:00. Bars before Asia open or after US close belong to no window and are excluded from classification.

Direction labels

For each session, the direction is derived from the change in that session's cumulative AVWAP between session open and session close. Below a ±1 ES point threshold the session is labeled FLAT and excluded from the conditional cells.

Confirm / disconfirm criterion

The current default is price-consolidation: for each bar of the later session (EU or US), after skipping the first hour, take sign(close − anchor_avwap). If ≥80% of those signs are on one side, label the session CONFIRM (matching anchor direction) or DISCONFIRM (opposite). Otherwise the day is "mixed" and outcome is None.

The legacy AVWAP-cross sign-flip criterion (used in earlier studies and accessible via the --legacy flag) tracked when the later AVWAP crossed the anchor AVWAP and stayed crossed for ≥30 minutes. It is less robust to coincidental touches and produced inflated rates on certain setups (e.g. the "Asia-Reversal Long" myth in §5).

Defense categorization

Only computed when EU outcome is CONFIRM and Asia direction is non-FLAT. During US, the price is checked against the cumulative Asia AVWAP:

• UNTESTED: US price never touches the Asia AVWAP line.
• HOLD: price touches the line and the US close is on the Asia-direction side of it.
• BREAK: price touches the line and the US close is on the opposite side.

Statistics

All confidence intervals use the Wilson score method (robust at small n and at extreme proportions). P-values are two-sided binomial tests against H₀ p=0.5 with continuity correction (normal approximation, fine for n ≥ 15). A setup is called robust when the Wilson 95% lower bound of the dominant side clears 50%.

Reproducibility

The analysis is fully reproducible. Data file: terminal/public/dev_history/ES_history.json (127,584 5m bars, ETH). Scripts: tools/audit/avwap_eu_asia_alignment.py (per-day classification), tools/audit/avwap_factorial_stats.py (factorial breakdown), tools/audit/avwap_eu_asia_sweep.py (criterion sensitivity sweep). Stdlib only — no external dependencies.