Last Friday I finished with two levels and no forecast: the low $4,300s, which had been tested and reclaimed during Fed week, and the $4,380–4,400 band that capped the post-hike rally. This week both of them answered. The top held as resistance. The bottom gave way.
What happened
Gold stalled just under $4,400 on 18 September and never got back there. Monday opened with a firmer dollar and spot slipped about 0.7% to around $4,346. On Wednesday the US flash PMIs came in hot, with the composite at 58.4 and services at 58.7, the fastest pace of business activity in more than five years, and the dollar climbed to a near two-month high. Fed officials lined up behind last week's hike; Governor Barr said further policy adjustments are "likely to be needed". Gold fell through $4,300 and closed the session around $4,290. On Thursday it stayed below the level, trading down into the $4,250–4,280 area, a two-week low.
Underneath the price, the rate path was being repriced. CME FedWatch put the odds of another 25 basis point hike on 28 October at roughly 70%, up from about 55% a day earlier. Oil fell early in the week, which should have helped bullion. It didn't. When a supportive input gets ignored, it tells you which variable is in control, and right now that variable is the Fed. Gold is down roughly 9% over the past month and about a quarter below January's record.
Why this one is different
Twice this month I've written here about candles that went through a level without taking it: payrolls on 4 September and the FOMC decision on 16 September. Both were spikes inside a scheduled release, minutes of thin liquidity followed by a move back through the level within hours.
Wednesday doesn't fit that pattern. There was no single release candle doing the work. It was a three-session grind lower that accelerated on data and was then confirmed by the people who set rates. And the part that matters most: price did not snap back. It closed below $4,300 and stayed below it the next day.
That is the first signature of a level being taken rather than swept. First signature, not proof. The rule I keep coming back to is that a break has to survive a retest.
What the retest will look like
Sooner or later price comes back to $4,300 from below. There are two outcomes, and both are informative.
If it gets rejected there, with sellers defending what used to be the buyers' ground, then ownership has changed and $4,300 is now resistance. That is the condition a breakout system waits for.
If it gets reclaimed and holds, then Wednesday was just a slower, deeper version of the same inventory clearing, and the level goes back to being support.
I don't know which it will be. I know what each one looks like, and I know that trading before the answer arrives is guessing.
The same level three weeks earlier, and what it cost
On 14 September the system sold a break of this same zone, around $4,280–4,290 on my broker's feed. The first short worked: +€24.14 in seven minutes. Then three more shorts went in near $4,282, and all three were stopped out between $4,289 and $4,292 within the hour: −€17.08, −€20.99 and −€26.55. Net for the day: −€40.48, just under 1% of the account.
The level wasn't taken that day. Price reclaimed it within the hour, slipped under it again only during the Fed release (exactly the kind of candle I wrote about last week), and by 18 September gold was pressing $4,400. The stops did what they are there for. Because they were set from the level and not from the candle, "wrong about the break" cost one percent instead of turning into a problem.
This week there was a single position, a short that was closed within a minute at +€1.71. It was a scratch, and the account has been flat since.
I show the 14 September losses because they are the honest half of the story. A level that finally breaks on the fourth test looks obvious in hindsight. It didn't look obvious on the third.
Where the record stands
September is the kind of month I warn subscribers about. The balance is down about 0.9% month-to-date. There have been three scheduled catalysts in four weeks, a level that faked before it broke, and very few clean setups. The profit factor across the full history has slipped from 2.22 at the start of the month to 2.14. That is what a choppy month does to a ratio built over two years, and I'd rather you see it here than discover it on the chart.
The full verified record, two years live on a real-money RoboForex account: 811 trades, +196.23% gain, 80% win rate, profit factor 2.14, maximum drawdown 15.32%, average 4.40% per month.
One note on numbers, because readers see two figures side by side. My MQL5 signal page shows +151% because MQL5 measures growth from the day the signal was published (26 February 2026, equity €1,664.59). The full track record shows +196.23% because it measures gain on deposits across the entire history since August 2024. Same account, two starting points.
Into October
I'm describing the environment, not forecasting. Friday's PCE release is the next scheduled catalyst, followed by the FOMC on 28 October with a hike currently priced at around 70%. Oil and the Strait of Hormuz remain an inflation wildcard in both directions. That is a calendar that will keep producing candles that look like breakouts.
The levels I'm watching: $4,300 overhead, now on probation as resistance. Above it is the $4,380–4,400 band that capped the post-Fed rally. Below is the $4,230 area and then the round $4,200 that many analysts treat as a baseline. Wednesday's close doesn't decide whether $4,300 has really been taken. The retest does.
You can follow AlphaEdge Smart with the full verified statistics on my profile. If you have questions about the retest condition, the stop logic or anything else in this post, ask in the comments. I answer everyone.
Past performance is not indicative of future results. Trading leveraged instruments carries a substantial risk of loss. Copying a signal introduces additional differences in spreads, slippage and execution that can materially affect results.
