Two of Our Forward Accounts Are Red. We Traced Every Trade: It Is One Signal Shown Twice

Two of the forward accounts on our public ranking page are red. ATLAS PORTFOLIO shows −10.48% (19 trades, profit factor 0.10, drawdown 11.53%). MEGAMAX EURJPY shows −13.89% (14 trades, profit factor 0.09, drawdown 15.22%). Both are demo accounts on the same broker's demo server, both started on 2026-08-21. We went through every trade, and this is what we found.

We took the deal journal from each terminal and paired entries with exits FIFO, which gives round trips. ATLAS: 21 round trips, 4 wins, net −960. MEGAMAX: 16 round trips, 3 wins, net −1,191.

ATLAS PORTFOLIO is a multi-pair portfolio EA. One of its components is a JPY-buy Donchian breakout, and that is exactly the logic MEGAMAX DONCHIAN TREND runs on its own. Laid side by side, the entries fell on the same timestamps on the same symbols on both accounts: Aug 25 03:00 EURJPY, Aug 28 09:00 USDJPY, Sep 14 13:00, Sep 16 00:00, Sep 18 03:00, server time. Two accounts, one signal. What the page shows as two losing EAs is the same losing leg displayed twice, not two independent failures.

Every loss was a 1R stop or a trailing-stop exit, about 1–2% of balance each. No oversized trade, no stop that failed to fire. Between Aug 24 and Sep 2, USDJPY fell from about 160 to 154, a yen-strength move, and the JPY-buy leg was stopped out nine times in a row. We checked the MEGAMAX source for a trend or regime filter: there is none (we grepped; zero hits). It buys the breakout and takes its 1R.

Our published backtest for this logic (2017–2026) has a win rate of 25% on EURJPY and 35% on USDJPY; the longest losing streak inside it is 17 on EURJPY and 11 on USDJPY. A streak of 9 sits inside that distribution. It is not evidence that the strategy is broken. It is also not pleasant: the forward account is what "inside the distribution" feels like in real time instead of in a report.

First, a portfolio EA that shares a leg with a standalone EA is not diversification against it. Run ATLAS PORTFOLIO and MEGAMAX DONCHIAN TREND on the same account and you hold the JPY-buy leg twice and lose twice on days like Aug 28. Do not run both. Second, read a streak against the published maximum streak, not against how it feels. Nine losses on a logic whose backtest shows 17 is the logic working as tested.

We will not change the strategy because of five weeks of demo data; that is the reaction the backtest exists to prevent. What we are considering is a backtest of a regime filter on the JPY-buy leg, for example only buying above a long moving average on H4. As of publication it is unresolved and untested, so there are no numbers and we are not claiming it would have helped. For context, the only clear winner on the same page in this period is our gold breakout EA, at 103 round trips, profit factor 4.41 and +15.28%.

The two EAs in this article: ATLAS PORTFOLIO https://www.mql5.com/en/market/product/182751 · MEGAMAX DONCHIAN TREND https://www.mql5.com/en/market/product/181838