Delta and cumulative volume delta (CVD) describe the current bar very well but predicted the next bar almost not at all in our tests. Trading every aggressive-volume marker or fading absorption lost money after costs. Where order flow did help was as a filter on a setup that already had an edge: skipping opening range breakouts whose CVD disagreed removed a group of losing trades.
Same bar vs next bar
On a few days of tick data, a one-minute bar's delta (buyer volume minus seller volume) correlated +0.80 with that same bar's price change: delta tells you who pushed price. Its correlation with the next minute's change was −0.06, essentially zero. By the time a bar closes, the information in its delta is already in the price.
Trading every aggressive marker
We tested buying and selling on every aggressive-volume marker (a big bar with heavy one-sided delta), with 9,720 combinations of timeframe, stops, trailing exits, targets and adding contracts on repeats, over 12.5 years plus a year of true tick delta. Before any costs, the profit factor was 0.95 to 0.97: no edge. Adding a contract on each repeat signal only multiplied the loss.
The few combinations that looked good all held the day's first marker to the close. That marker was usually the 09:30 opening bar, and simply trading the opening bar's direction without any order flow did just as well. The markers look accurate on a chart because they print on the biggest bars of each move, which is hindsight.
Absorption fades
Absorption is heavy volume into a level that fails to move price. Fading it toward the next key level was real in risk units, +49.6R over three years, but it won only 22% of the time over 964 trades with tight stops. After a $2.50 round trip per MNQ trade, three years made $204.
Where it helped
- As a confirmation on a breakout: requiring CVD to agree with an opening range breakout removed trades that lost money in both of our test periods.
- As context: delta divergence and absorption tell you who is trapped, which is useful when you already have a level and a plan.
Simulated results on historical data, after commissions. They don't guarantee future results; see the risk disclosure and CFTC Rule 4.41.