I Lost 96% of a Live Account. Here Are the Five Checks I Now Run on Every EA, Including My Own.

I published an expert advisor that lost almost its entire account. The record is still on my profile. This post is what I learned from it, written as five checks you can run on any EA before you trust it with money — including mine.

The account I did not delete

One of my published live accounts is down roughly 96%. Different EA, different strategy, different codebase to the one I sell now, but mine, and still visible.

I left it up for a practical reason rather than a noble one. If I delete the accounts that failed, every account I keep becomes worthless as evidence. A developer whose published history contains only winners is not a developer with only winners. It is a developer who deletes.

What that account actually cost me was not the balance. It was eighteen months of believing that a good entry signal was the hard part.

It is not. The entry is the easy part. Position sizing, stop placement and the rules that decide when the system is not allowed to trade are the hard part, and they are the part that determines whether an account survives a bad month. I now build that layer first and fit the entry logic to it afterwards.

Below are the five checks that came out of that. They are not specific to my software. Run them on anything you are considering, mine included.

Check 1: Does every position carry a broker-side stop loss?

How to check it. Run the EA in the Strategy Tester, open the Results tab, and look at the S/L column on the entry rows. Every row should have a price in it.

What a failure looks like. An empty S/L column. The seller will often describe this as a "virtual stop", "smart exit" or "adaptive close", meaning the stop exists only inside the EA's own logic.

Why it matters. A virtual stop is not a stop. It requires your terminal to be running, your VPS to be up, your connection to be alive and the EA to still be responding. A broker-side stop sits on the broker's server and executes when none of those things are true. The difference shows up exactly once, on the day it matters, during the kind of move that also takes your connection down.

This check on its own eliminates a large share of what is on the market.

Check 2: Does the lot size increase after a loss?

How to check it. In the tester results, sort by time and read the Volume column down the page. Note what happens to volume immediately after each losing trade.

What a failure looks like. Volume rising after a loss, then rising again after the next one. 0.01, then 0.02, then 0.04, then 0.08. That is martingale, whatever the description calls it. Related patterns: several positions opening on the same symbol at worse and worse prices, which is grid or averaging.

Why it matters. Martingale produces a beautiful equity curve almost all of the time. That is the entire problem with it. It converts a large number of small wins into one loss that takes everything, and the backtest will usually look excellent right up until the sequence that ends it. A curve with no visible drawdown is not evidence of low risk. It is frequently evidence that the risk has not arrived yet.

Be aware that a martingale EA can pass Check 1. Stops that are real but sized to a doubling sequence still end the account.

Check 3: What did the return cost in drawdown?

How to check it. Take the total return and divide it by the maximum drawdown. That is the recovery factor, and the report gives it to you directly.

What a failure looks like. A large return sitting next to a large drawdown. 200% return with 60% drawdown is not a strong result. It is a leverage setting. Halve the risk input and you have roughly 100% return with 30% drawdown from the same code.

Why it matters. Return on its own tells you nothing, because return is adjustable. Anyone can double their headline number by doubling risk, and many published results are exactly that. The ratio is the part that describes the strategy rather than the settings.

Ask yourself the practical version of the question: could you actually sit through the drawdown figure without switching the EA off? A drawdown you will not tolerate in practice is a drawdown you should not run, no matter what the return next to it says.

Check 4: How large is the sample, really?

How to check it. Count the trades, not the months. Then check whether the profit is spread across them or concentrated in a handful.

What a failure looks like. Forty trades presented as a proven system. One trade producing a third of the total result. Ten months of flat performance and one enormous month carrying the whole curve.

Why it matters. Forty trades cannot separate skill from luck. Neither can a hundred, honestly, though it is better. This applies to my own results as much as anyone's, and I will apply it to them at the end of this post.

There is a specific version of this worth checking on live accounts: look at the profit broken down by symbol. If a multi-symbol EA made all of its money on one symbol, you are not looking at a multi-symbol system. You are looking at a single-symbol system with some other charts attached.

Check 5: Does the track record belong to the version being sold?

How to check it. Compare the date the monitored account opened against the version history of the product.

What a failure looks like. An account opened three years ago attached to a product currently on version 12. That curve was produced by versions 1 through 11, by code that no longer exists. The version you would be buying has contributed the last few weeks of it, and you cannot tell which weeks.

Why it matters. This is the most common way results are technically true and practically meaningless. Nobody is lying. The account is real and the trades happened. They were simply produced by different software.

My own answer to this is to open a new live account each time I release a version, fund it from zero, and let the previous record end where the previous version ended. It makes my numbers look smaller and shorter than they could. It also makes them mean something.

Running all five on my own EA

It would be dishonest to publish a test and then not take it. Apex Drawdown Zero V9, on the live account documented in my previous post:

Three clear passes, one clear fail, and one that needs stating plainly rather than scoring: applying Check 4's symbol test to my own account, gold produced +$290.98 while EURUSD finished at -$5.01 and EURJPY at -$20.68. Over that period the profit came from one symbol. I would rather tell you that than have you discover it and wonder what else I left out.

So: a system that passes the structural checks, on a sample too short to prove anything yet, currently carried by one symbol. That is an accurate description, and it is the one I would want if I were the buyer.

What to do with this

Take the five checks and run them on whatever you are looking at right now. Do not start with mine. Start with the one you were already close to buying, or the one already running on your account.

Most of the time Check 1 and Check 2 are enough, and they take about four minutes in the Strategy Tester. If an EA fails either of those, nothing in its description matters.

The demo of my own EA is free on its product page, so you can run all five on it without spending anything, and compare what your own tester produces against the live figures I published. If it does not hold up under your testing, do not buy it. An EA you do not trust gets switched off during its first bad month, which is the worst possible moment and a worse outcome than never running it.

If you run these checks and find something that surprises you, message me through MQL5 and tell me. I would like to know, including if it is about mine.

Tshivhidzo Moss Mbedzi, MPC Labs, South Africa

Risk disclosure

Trading foreign exchange and commodities on margin carries a high level of risk and is not suitable for every investor. Past performance does not guarantee future results and is not indicative of future performance. The figures referenced cover 42 days and 43 trades, which is a short sample and is presented as one. Nothing in this post is a projection, forecast, promise or guarantee of any future result. Any expert advisor can lose money, including mine. The checks described reduce certain categories of risk; they do not eliminate risk and they cannot prevent losses caused by market gaps, slippage, broker execution or connectivity failure. Do not trade money you cannot afford to lose. Nothing in this post is investment advice.