MidasA stress test runs your exact grid settings through the hardest stretches of real gold history — such as 29 January 2026, when gold fell $314.21 within one hour — and answers three questions: did the account survive, how deep did the deepest basket go, and how much deposit would it have needed; the free Midas Grid emulator does this day by day on real ticks since January 2025.
Results of a model on past ticks, not a forecast and not advice. Trading with leverage carries risk: if the deposit is too small, a stop-out is possible.
On this page (10)
What is a stress test, and how is it different from a backtest?
A backtest runs settings over a period and reports the result: profit, drawdown. A stress test is a backtest aimed at the worst cases on purpose: the widest hours, moves without a pullback, several feeds with slightly different prices, and a deposit that is not topped up. For a martingale grid the stress test matters more than the average, because a grid usually earns little and often and loses rarely and a lot.
Real ticks means every price change the market printed, not prices rebuilt from one-minute candles. A grid lives on dollar distances, and within a minute gold can travel $20, so candles hide exactly the moves that fill a grid.
Step 1. Do the arithmetic before the test
Enter your deposit, first lot, step, number of orders, leverage and stop-out level in the grid calculator. It shows the stop-out distance: how far gold has to go against the first order to close the account.
Compare it with ordinary gold. In 2026 the median day moved $101 from high to low (gold daily range). With the calculator defaults ($1,000, 0.01, ×2, step $1.95, 11 orders) the stop-out is $65.58 away. If your distance is below a typical day, you already know the answer; the test will only show you the date.
Step 2. Set up two variants in the emulator
- Open the Midas Grid emulator and switch to "Settings test". It is free and runs in the browser, no MT5 or VPS needed.
- Variant A is the reference (Midas 5.41: $1,000, lot 0.01, 11 orders, step $1.95). Leave it as is.
- In variant B enter your own capital, first lot, number of orders and step, plus your broker's spread, leverage and stop-out level.
- Tick "Start over after a blow-up", so the test counts every blow-up instead of stopping at the first.
The emulator models a doubling grid (×2) of the Midas type; it does not run other expert advisors.
Step 3. Pick the whole history, not a good month
Choose the period "All" (since January 2025). A month or a quarter can easily miss every hard stretch. Periods of up to a month are open without signing in; the full history opens with a free code from the Telegram bot.
The stretches the test must include:
| Hour (UTC) | Gold in that hour | Why it matters |
|---|---|---|
| 29 Jan 2026, 15:00–16:00 | −$314.21 (range $397.29) | The widest hour of the data; a full basket on the wrong side has no chance |
| 30 Jan 2026, 17:00–18:00 | −$135.62 | The next day, the move continued |
| 2 Feb 2026, 05:00–06:00 | −$130.58 | Monday after the weekend |
| 5 Feb 2026, 02:00–03:00 | −$147.67 | Asian hours, no scheduled release |
| 3 Mar 2026, 14:00–15:00 | −$118.58 | A second wave a month later |
| 23 Mar 2026, 00:00–01:00 and 11:00–12:00 | −$110.92, then +$110.06 | Both directions on the same day |
16 of the 25 widest hours of 2025–2026 fall between 28 January and 2 February 2026, and none of the 25 had a scheduled US or ECB release. The full table: biggest 1-hour gold moves.
Step 4. Read the result in this order
- Blow-ups. The count comes first. One blow-up is one lost deposit; profit before it does not matter.
- Seven feeds. Tick "Spread right away: 7 feeds": each variant also runs on 6 feeds with prices ±1 cent and a different execution delay, like different brokers. If a variant blew up on even one feed, it can do the same at yours. Look at the worst case, not the best.
- "Min capital, no blow-up". The deposit these settings would have needed over this period. If it is above your capital, the settings did not pass. The emulator cannot compute it when a basket stop or profit withdrawal is on.
- "Max drawdown". The deepest fall from a peak. Ask whether you would have sat through it on a real account without switching the bot off.
Step 5. Replay the hardest days tick by tick
Open the "Replay" tab, pick 29 January 2026 (or any day from the table) and play it. Use "⏭ basket" to jump to the close of each basket and read the basket journal: how many orders it took, how deep it went, how it closed. Watch where the stop-out line sat compared with the move. This is where a number becomes a picture you remember.
Step 6. Change one setting at a time
Go back to "Settings test" and change one thing in variant B:
- the number of orders (11 → 10 → 9), which halves the basket at the bottom each time;
- the deposit per 0.01 lot;
- "Basket stop, % of balance" together with "Pause after stop" (see 7 protections);
- the step (see ATR grid step).
Run each change over the same whole period and write down blow-ups, worst feed and minimum capital. Two changes at once will not tell you which one helped.
Step 7. Decide, then confirm on demo
Settings pass the test when, over the whole history: no blow-ups on any of the 7 feeds, the minimum capital is below your deposit with room to spare, and you could live with the deepest drawdown. Then run the same settings on a demo account at your own broker for a few weeks. The broker's spread, swap, execution and stop-out rules are what the model does not know.
What the test cannot tell you
- The future. The next crisis will not repeat 29 January 2026. Passing the past is a filter, not a promise.
- Your broker. The model uses a constant spread, no slippage and no requotes.
- The size of the next move. A deposit that survived the worst hour of 2025–2026 might not survive a wider one.
We treat our own model the same way: a result in the model is a reason to test on demo, not a verdict.
FAQ
How long should a stress test period be?
The whole available history, since January 2025 in the emulator. A short period can miss every hard stretch; late January 2026 alone holds 16 of the 25 widest hours of the data.
Which days should I always include?
29 January – 2 February 2026, 3 March 2026 and 23 March 2026. The first holds the widest hour of the data (−$314.21 between 15:00 and 16:00 UTC on 29 January); on 23 March gold moved more than $110 in each direction.
Is the MT5 strategy tester enough?
It can test on real ticks if your broker provides them, and it runs any EA. It does not show seven feeds at once or the minimum capital without a blow-up; the emulator does both for the Midas grid. Using both is better than either.
If my settings pass, is the account safe?
No. Passing means the settings survived the past in a model. A future move can be wider, and a broker adds spread, slippage and its own stop-out rules. Confirm on demo and keep a reserve.
Can I stress test another EA in your emulator?
No. It models the Midas grid (×2 lots, your step, orders, deposit and stops). For another EA use its own tester on real ticks with the same worst days.