- Is Cloudbank a martingale or grid EA?
- No. It opens one position at a time with a fixed lot, a fixed 70-pip stop and a fixed 140-pip target. There is no averaging down, no grid, no recovery trade and no lot progression of any kind. A crossover that happens while a position is open is ignored. The maximum number of positions open at the same time, measured across the whole published record, is one.
- How was it tested?
- On Exness MT5 USD/JPY history in the MetaTrader 5 strategy tester, on the "every tick based on real ticks" model (Model 4), from 2019-01-01 to 2026-07-01 with 0.1 lot on a $10,000 balance. The tester reported the history quality of that run as 100% real ticks over 229 million ticks. All 646 closed trades are published on this page and downloadable as CSV.
- Does the backtest include swap?
- Yes, and it is not a rounding error here. The trades made 4,355.52 before carry; swap subtracted 800.57 of it, and the published net is the figure after that. Roughly 18% of the gross went to carry, because 469 of the 646 trades stayed open across at least one night and the median hold is about 25 hours. Commission was zero in this run, so a broker charging per-lot commission will subtract more.
- Why does it lose more trades than it wins?
- Because the target is twice the stop. Risking 70 pips to make 140 does not need to be right half the time — it needs to be right often enough that the larger winners outpay the smaller losers. Over the published record the average win was 49.37 and the average loss 36.03. A high win rate is not what this EA is built to produce.
- The verification window is negative — what does that mean?
- It means 29 trades since the 2026-04-01 cutoff came to -83.05, a profit factor of 0.86 on a maximum drawdown of 1.99%. Twenty-nine trades is still too few to conclude anything. For scale, the measured backtest contains single months down 415.14, 234.20 and 192.60, so a quarter like this one sits inside the range the strategy has already shown. The figure is published as measured rather than smoothed, and the window will fill as more trades are recorded.
- Does it trade at particular hours?
- No. There is no time, session or day filter. Any closed H1 bar on which the conversion line crosses the base line can open a trade, provided no position is already open.
- How often does it trade?
- About seven times a month across the published record — 646 trades over seven and a half years. The rate is not constant — the busiest year in the record carried nearly three times as many trades as the quietest one. It depends on how often USD/JPY changes direction, which is not something a setting can turn up.