baseedge_labs
← research

Eight Experiments on What Actually Works Around VWAP — and What Only Looked Like It Did

Aug 25, 2026 · BaseEdge Labs research

Drift VWAP

Eight separate experiments on CME Micro Nasdaq-100 futures, January 2024 through August 2026. Results are hypothetical backtests over that period.

What we found

  1. The band-fade edge does not survive realistic execution. It looked like the standout in a frictionless study, then lost money out of sample once fills, slippage and commission were modeled. This is the most important finding in the lab.
  2. ATR-relative sizing helps London, not Asia or NY. It turned London’s losing in-sample window into a profit and cut full-history drawdown sharply. Asia and NY did better on their existing fixed-point brackets.
  3. Tuned Efficiency Ratio settings improved the full history, but the gain came from the window they were tuned on. They cut a slice of low-probability trades. Out of sample they earned less than the shipped baseline.
  4. On MNQ, ATR did not trend mechanically with price level over this window. Long bull runs often came with compressed ATR. A 10-point floor on ATR-relative brackets does real work in quiet sessions.
  5. The shipped 45+ point Asia target was fitted to a high-volatility stretch. In early 2024 few Asia trades ever reached 40 points. A grid found configs that were net positive in every 4-month chunk.
  6. The drift-retest entry works best in Asia and London, and worst in NY. NY whipsaws and needs extra filtering.
  7. Most fresh 5-minute VWAP crosses pull back to VWAP before they run. That argues for a staggered entry: a small probe on the cross, the rest at the retest.
  8. NY needed the most filtering of the three sessions across nearly every experiment here.

Experiment 1 — ATR-relative sizing vs. the static-points baseline

Method. We swept the v4.0.0 per-session ATR-multiplier inputs in-sample (2024-01-01 → 2026-05-10, 900 combos per session). The top performer was then checked out of sample (2026-05-10 → 2026-08-22, holdout) against the shipped v3.0.1 static-point brackets.

Findings.

Recommendation: switch London to ATR-relative sizing; leave Asia and NY on points.

Experiment 2 — Confluence filter tuning (Efficiency Ratio & MACD Velocity)

Method. We swept Kaufman’s Efficiency Ratio and MACD Histogram Velocity in-sample, to see whether tighter gating removes false entries without over-filtering real ones.

Settings found: London ER period 9 / threshold 0.15 / fade-drop 0.15 (with MomVel active); NY ER period 4 / threshold 0.05 / fade-drop 0.15.

Findings.

Experiment 3 — MNQ ATR(14) characterization

Method. We compared MNQ price level with raw-point ATR(14) bar by bar, and split the trading days into four volatility quartiles.

Findings.

Experiment 4 — Exploratory VWAP entry mechanics (frictionless)

Method. An initial, theoretical forward-excursion study across several VWAP-based entry types: band fades, momentum continuation, liquidity-sweep reclaims and break-and-retest. It measured pure bar-level MFE/MAE at a 1.5R target, with no execution model at all.

Findings.

See Experiment 5 for why these readings did not hold.

Experiment 5 — Realistic execution check on Experiment 4

Method. We re-ran the standout band-fade and liquidity-sweep findings under real execution assumptions:

Findings.

Verdict: do not build a standalone band-fade module into the shipped strategy. A theoretical edge that disappears once friction is applied is worse than no edge at all, because it would have shipped.

Experiment 6 — Asia session, 4-month chunks

Method. To find out why the shipped Asia config drew down in 2024, we split the window into eight 4-month chunks. In each we measured the share of Asia trades that reached 25, 40 and 50 points. We then ran a 1,560-combination grid: TP 10-50 points, SL 25-100 points, VWAP slope 0.5-2.5 points, ER filter on/off and threshold, and MACD velocity.

Findings.

Experiment 7 — VWAP slope expansion & drift-retest mechanics

The mental model: VWAP’s slope is a lagging, volume-weighted reaction to price that has already printed. Going short only because price already fell hard and VWAP slopes down means selling into an extended zone. Institutional liquidity tends to absorb the move there. That is the “chasing expansion” trap. The edge sits on the first minor pullback toward the sloping VWAP (0.2-0.8x ATR away), entered in the direction of the confirmed volume trend.

Method. We measured drift-retest signals on active VWAP down-slopes on NQ, per session. For each we recorded the ratio of favorable to adverse excursion (MFE/MAE), and the result of a fixed 1.5x / 0.8x ATR bracket.

SessionEdge ratio (MFE/MAE)1.5x / 0.8x ATR bracketBehavior
Asia (18:00-03:00 ET)FavorableNet positive; strongest of the threeHigh continuation persistence — overnight drift holds direction
London (03:00-09:30 ET)FavorableNet positiveStrong directional expansion — momentum runs
NY (09:30-16:00 ET)Unfavorable (MAE outran MFE)Net positive; weakest of the threeFrequent whipsaws and re-crosses — needs extra filtering

Findings.

Experiment 8 — First 5-minute VWAP cross (“trend inception”) scalping

Method. Does entering on the very first 5-minute candle that closes across VWAP work, without waiting for a pullback? We tested raw, unfiltered signals with two brackets, +30 / -20 and +40 / -20 points. We then added an “expansion energy” filter: volume above 1.2x its 20-bar average, or Kaufman ER above 0.30.

Findings.

Caveats

Consolidated recommendations

  1. Do not build a standalone band-fade module. Experiment 5 showed it fails realistic execution.
  2. London: switch to ATR-relative sizing (2.0x / 1.0x drift-pullback bracket, or the swept 4.0x / 5.0x). Asia and NY: keep fixed points, but revisit Asia’s target given Experiment 6: the 45+ point targets were tuned to an unrepresentative high-volatility stretch.
  3. Use a staggered entry model. Half size on the inception cross (gated by volume or ER), half size as a limit order at the VWAP retest (0.3x ATR). It catches both the minority of runaway moves and the majority that pull back first.
  4. NY needs the most filtering of the three sessions across nearly every experiment in this lab. Gate it with Efficiency Ratio and MACD Velocity rather than trading it on raw VWAP conditions alone.