Fading the top and bottom of a detected trading range lost money on NQ in every form we tried. The rule can't tell a range that will hold from one about to break, and every breakout stops the fade out. The consolidation trade that did hold up was the opposite: the breakout from the opening range.
How we detected a range
Over the last 40 bars: price going nowhere (a low ratio of net move to total path travelled), a width within normal limits, and several touches near both the top and the bottom. The trade bought a poke into the bottom that closed back inside and sold the mirror at the top, with a stop beyond the edge and a target most of the way across.
Results
On one-minute bars the rule fired 15,758 times in three years with a profit factor of 1.00 in risk units. After costs it lost the whole test account. On 5-minute bars it lost 164R. We then tested 36 variants of lookback, entry depth, target and minimum width: the best had in-sample profit factors of 1.00 to 1.03, and none was profitable on the later period it hadn't seen.
Bounding the range with meaningful levels instead, the prior day's value area, didn't change the verdict either (see the value-area study).
Why
A range only looks like a range until it breaks. A human can often see a clean, well-tested range and watch the order flow at the edge; a rule sees two touches and a low efficiency ratio, and takes the trade every time, including right before the breakout.
What works with consolidation
The breakout. The first 15 minutes of the session is a range every trader sees, and trading its break with the stop on the far side held up in both of our test periods once it was filtered for the day's direction and order-flow agreement. (Without those filters it only worked from 2021 on.)
Simulated results on historical data, after commissions. They don't guarantee future results; see the risk disclosure and CFTC Rule 4.41.