MR4H · V5.3

Mean-reversion basket,
overnight swing

Three sleeves — EURUSD, GBPUSD and US500 (long-only) — merged into one basket. The system buys extreme stretch back to the mean, with a trend filter and a sweep condition. Entries cluster in the NY afternoon; 82% of positions are held through one night.

How it works

The structure is public; exact threshold values stay private.

1

Extreme stretch

A short-period RSI at an extreme reading and price stretched a multiple of ATR away from its short-term mean — double evidence that the market has overshot.

2

With the long-term trend

A 200-period trend filter sets the allowed direction. US500 trades long-only — shorts proved to be a structural drag on a rising index.

3

Sweep condition

The stretch must coincide with taking out a recent extreme — the same liquidity logic as the OK System, but on the 4H chart.

4

Exit back at the mean

Profit is taken as soon as price tags its mean again; stop at 2×ATR, maximum hold 20 bars (4H). Median 16 hours in the trade, at most ~3.3 days.

Key numbers per sleeve

2017 – 2026 · 1% risk per trade each · costs (~1.1 pip) baked in.

SleeveTradesResultPFWinMax DD
EURUSD80+14.3R2.3980%−2.3R
GBPUSD79+16.0R3.0577%−2.7R
US500 · long85+9.6R1.6372%−3.4R
Basket244+40.0R2.1976%−3.9R
The diversification gain is the key finding: the sum of the individual max drawdowns is −8.5R; combined it is −3.9R. More than halved, purely by merging sleeves that don't share a monthly rhythm.

In-sample vs. out-of-sample — the firewall

The configuration was chosen on 2023–2026. Everything before that was tested blind: that data never influenced the parameter choice.

PeriodResultPF
2017 – 2019out-of-sample · blind+11.8R1.92
2020 – 2022out-of-sample · blind+9.6R1.98
2023 – 2026in-sample · selection period+18.6R2.70
All OOS combined151 trades, blind+21.4R1.95

Read it like this: the in-sample number (PF 2.70) is flattered by definition — that's what was selected on. The honest number is the blind period: PF 1.95 across 151 trades, including Brexit, COVID and the 2022 shocks. That the system survived those without prior knowledge is the real evidence.

Why the sleeves don't bite each other

Spot prices of EURUSD and GBPUSD correlate ~0.90. This system's trades don't.

EU ↔ GU
+0.12
monthly R-correlation
EU ↔ US500
−0.11
slightly negative
GU ↔ US500
−0.11
slightly negative

The system picks different moments in each market — which makes the basket far more robust than the spot correlation suggests, and makes US500 a genuine diversifier.

Time-in-trade profile

Short-swing / overnight — not intraday, not long-term.

< half a day
16%
within 1 day
58%
1 – 3 days
26%

Median 16 hours, average 24 hours, 82% held overnight. Entries cluster in the NY afternoon; exit is usually the next morning. That's why MR4H doesn't clash with the OK System in either time or horizon (London morning, flat the same day) — exactly why the two can form a 6-sleeve portfolio.

Live caveat: positions stay open through scheduled news (NFP/CPI/FOMC/ECB). The 2017–2026 backtest spans Brexit, COVID and 2022 and survived — but a tier-1 news gap can gap past the stop live. A news filter is on the research list.