The London Reversal: a session playbook with 21 years of M1 data

The London Reversal is the most-discussed session pattern in retail FX. Every trading forum has a thread on it. Every TradingView indicator author has an opinion. The problem is that almost every published test of it runs on Dukascopy M1 — which, as we covered in the previous post, isn't enough.

This post documents what we found when we ran the London Reversal on Deepwick's full-tick M1 tape from 2003 to 2024.

The setup

The London Reversal entry, in its simplest form, is:

1. The London open (08:00 BST / 07:00 UTC during winter) sweeps the prior Asian session high or low.
2. The sweep fails — price returns inside the Asian range within 30 minutes.
3. Enter in the direction of the failure (short if the high was swept, long if the low was swept).
4. Stop beyond the sweep extreme.
5. Target: midpoint of the Asian range, or full range, depending on version.

We tested six variations on this pattern across EURUSD and GBPUSD, with the following constants:

• Asian range defined as 00:00–07:00 UTC (the most common retail definition)
• Sweep threshold at least 5 pips beyond the prior high/low
• Entry window 07:00–08:30 UTC
• Stop 10 pips beyond the sweep extreme
• Target full retracement of the sweep (i.e., back to the prior Asian extreme)

We ran the test on Deepwick's full-tick M1 from 2003 to 2024, then repeated on Dukascopy M1 for the same window. Both tapes are sampled at M1 resolution; the difference is the source aggregation.

What full-tick M1 shows

The headline numbers are striking. Across the 21-year window on full-tick M1:

• EURUSD long (low sweep): 58% win rate, +0.42R average per trade, profit factor 1.6
• EURUSD short (high sweep): 53% win rate, +0.21R average per trade, profit factor 1.3
• GBPUSD long: 56% win rate, +0.31R average per trade, profit factor 1.5
• GBPUSD short: 49% win rate, -0.05R average per trade, profit factor 0.95

The asymmetric edge — long on the low sweep, particularly on EURUSD — is consistent with what retail traders have reported anecdotally for years. What was less expected was how much the numbers degraded on Dukascopy M1 for the same window:

• EURUSD long (Dukascopy): 51% win rate, +0.04R average per trade, profit factor 1.05

Same window. Same entry logic. The Dukascopy tape turns a clearly profitable pattern into a marginal one.

Why the gap?

Two causes, both of which we covered in the previous post:

1. Wick undercount. Dukascopy's tape systematically undercounts the London open wicks. The pattern's whole edge depends on those wicks being captured and exceeded. When they're smoothed by missing tape, the sweep looks shallower than it actually was live — your stop is tighter than it should be, your win rate drops.

1. Spread regime distortion. The London open spread on the real interbank tape spikes much harder than on a single broker's tape. Strategies that assume a flat 1-pip spread through the entry window are modelling a calmer market than they'll actually face.

These two effects compound. The strategy isn't just losing on individual trades; it's losing more on each losing trade and winning less on each winning trade.

The overfitting test verdict

We ran the optimisation across 6 variations × 4 instruments = 24 strategy candidates. The Deepwick overfitting test reported:

• EURUSD long — robust (all three signals pass)
• GBPUSD long — robust
• EURUSD short — caution (plateau narrower than ideal)
• GBPUSD short — likely overfit (forward window failed)

This matches the raw numbers. The long-side London Reversal on EURUSD and GBPUSD has a real, persistent edge across 21 years of full-tick data. The short-side versions don't.

If we'd run this on Dukascopy M1, we'd have concluded none of the variants worked — and missed a real pattern that's been tradable for two decades.

What this changes

Three things, all of which are downstream of having a better tape:

1. Stop placement. With full-tick wicks, the right stop for the EURUSD long version is 12-14 pips beyond the sweep extreme, not the 8-10 pips a Dukascopy-only backtest would suggest. The 8-10 pip stop gets taken out by wicks you'd never see on Dukascopy.

1. Entry filter. The high-quality entries on the EURUSD long version happen when the Asian range is wide (over 40 pips for EURUSD). Tight Asian ranges produce marginal entries even on full-tick data.

1. Holding period. Full-tick data shows the target is usually hit within 90 minutes, not the 30-60 minutes Dukascopy suggests. The pattern takes longer because the wicks take longer to retrace than the smoother Dukascopy tape implies.

The honest disclaimer

A 21-year backtest on full-tick M1 doesn't prove the London Reversal will work for the next 21 years. Markets evolve. Liquidity providers rotate. The pattern could fade.

What it does prove is that the pattern exists in the data traders actually face. That's a different and stronger claim than what a Dukascopy-only backtest supports.

We'll publish the strategy on the Deepwick marketplace when the platform opens, with the overfitting test verdict attached. Buyers will know exactly what they're buying: a 21-year-tested pattern with documented edge, transparent limitations, and explicit out-of-sample validation.

That's the whole point. The strategy either has an edge or it doesn't. The data should tell you which, not the marketing.

— Enrique