forex-scalping.com

Mean reversion at 3 a.m. and why the range keeps breaking

During quiet overnight hours, many scalpers try fading moves in range-bound pairs. It looks tidy: price oscillates inside a band, each touch of the edge produces a small bounce.

The strategy collects small wins at a high rate. It also carries a tail risk that the statistics hide until the day it appears. A single directional move, driven by an overnight headline or a large order, breaks the range and runs.

The positions built along the way, each justified by the previous bounce, all go wrong together.

This shape of payoff, many small gains and rare large losses, is characteristic of range fading in thin markets. The win rate can sit at 80 percent for months. Then one event takes back half a year of profit.

Defences include a hard cap on the total number of positions, an exit rule for when the range breaks, and avoiding times of the week when news is more likely.

Some check the news calendar even for the quiet hours, since data from Australia, New Zealand, and Japan arrive then.

What is the largest single loss in the strategy's history compared with the average win?

Price on a tick chart often looks like it reverses constantly. Up a bit, down a bit, up again. Part of this is a statistical illusion called bid-ask bounce.

If trades alternate between hitting the bid and lifting the ask, recorded prices jump between two values even when the true price hasn't moved. Negative serial correlation appears: an up tick is followed by a down tick more often than chance would suggest.

A backtest on trade or quote data that doesn't account for this finds huge reversion edges. They look wonderful, and they can't be harvested, because capturing them requires trading at the best price on both sides. Retail traders pay the spread.

There's a test. Compare the strategy's results using bid prices only versus filling buys at the ask and sells at the bid. If the edge collapses, it was the bounce.

Genuine mean reversion exists, but at the level of seconds it hides inside the noise and under the cost of crossing the spread.

Anyone found tick-level reversion that survived honest cost accounting?

In the quiet hours of the late Asian session and before the London open, many major pairs trade in narrow ranges with frequent small reversals. That has made mean reversion strategies popular for this period, with traders buying near the bottom of the range and selling near the top. The approach can work because there is little news and fewer large participants pushing price in one direction. The challenge is that spreads are often wider at this time, ranges are small, and the occasional breakout can produce losses large enough to wipe out many small gains. Careful cost analysis is essential before relying on this kind of strategy.

I ran a range-fade on EURCHF and AUDNZD in the early Asian hours for about five months, and the win rate sat near 78 percent the whole time. What nobody tells you is how the one bad day looks. A central bank headline came out around 4 a.m. my time, the range broke by eighteen pips, and I had four positions stacked along the way, each one added because the previous bounce had worked.

That single morning erased roughly seven weeks of profit. After that I capped the basket at two positions, put a hard exit ten pips beyond the range edge, and refused to add if the spread was above one pip. The win rate stayed about the same, but the worst day shrank to something I could sleep through.

I also started checking the calendar for any speaker scheduled within six hours of the session. If there was one, I simply stood aside. Boring, yes, but the equity curve has been much calmer since I made that rule absolute.