I've been porting a sweep → CISD → C3 model to MT5 for gold, and I ran three backtests on it. Here's what they said.

Run A — 12 trades, one month. 41.67% win rate. +$247.82 on $2,000.

Run B — 7 trades, one month. 85.71% win rate. 6.59 profit factor. +$466.10 on $5,000.

Run C — 223 trades, nine months, 100% history quality. 46.64% win rate. 1.44 profit factor. +$5,764.82 on $5,000.

Run B is the one that would sell. "85% win rate" on a sales page converts. It is also seven trades, which is not a result, it's a coin landing the same way a few times. The same EA that produced it produced a 41% month immediately next door.

So the number on my page is 46.64%, from the 223-trade run, and the page says out loud that more than half of all trades lose money.

Here's the part that took me a while to accept: that's the more persuasive number, not the less. A trader who understands liquidity models reads 46% with a 1.65-to-1 payoff and immediately knows what they're looking at. They can size it. They know what a bad week feels like in advance. Whereas 85% sets an expectation nothing can meet, and the buyer who believed it is refunding by week three and telling people the EA is broken. It isn't broken. The marketing was.

The bit almost nobody publishes: the worst stretch inside that profitable 223-trade run was eight consecutive losses and a 20.74% equity drawdown. If a buyer doesn't know that going in, they'll switch the system off at the exact worst moment, or double their lot size trying to win it back. Those two reactions destroy more accounts than bad strategies do.

---

The model itself, briefly, in case it's useful to you whether or not you ever touch my EA:

A higher-timeframe candle pushes past the previous candle's high or low, then closes back inside its range. Liquidity taken, immediately rejected. That's the sweep, and the close-back-inside is the part people skip.

Then you drop to your entry timeframe, find the exact bar that printed the sweep's extreme, and trace back through the unbroken run of candles that drove into it. The open of the oldest candle in that run is your level. Break it with a close and the move that created the sweep has been undone.

On the next higher-timeframe candle, one of two things happens. A close through that level, and you're in. Or price takes out the sweep's extreme first, and the setup is dead — no trade, no revenge entry, wait for the next one. Most sweeps never become trades. That's the model working, not failing.

---

I've written the whole thing up as a free PDF and attached it to this post. Full model breakdown, all three backtest numbers including the ones that don't flatter me, and an honest list of who the model suits and who it doesn't. No email needed, it's just attached.

If you build EAs: publish your drawdown and your worst losing streak next to your returns. It costs you a few impulse buyers and saves you every refund and every one-star review from someone who was never the right customer.

Backtest figures are MT5 Strategy Tester simulations, not live returns. Past performance does not predict future results. Trading leveraged gold CFDs carries substantial risk of loss. Not financial advice.