Gold after a sweep of yesterday's high: the money sits on Fed afternoons

Our coin-flip study found one basic idea on gold that beat a coin flip at the same bars: a rejection of yesterday's high or low. On the platform, the short side looked much better inside an event window, the minutes around a scheduled US release. We wanted to know whether the news does the work, or something else that happens at the same time.

• Gold futures, 1 Sept 2025 to early Nov 2026, on 1-minute and 5-minute bars. One micro contract, costs included.
• The trigger: price trades above yesterday's high, then the short. The stop and target are fixed and sized on recent volatility. No other filter.
• Each trade is tagged by whether a scheduled US release was near and by time of day: the first hour, the Fed afternoon, and the rest of the day.
• Luck check: each session takes another session's results at the same minute, 2,000 times. Because we chose the event window after looking, the honest test lets the luck side pick its best filter from the same list as well.

Over all shorts the idea was mildly positive. Inside an event window it was strong: about three quarters of trades won, against barely break-even outside. It held in both halves of the year, and only 1 of the 12 months that had window trades lost money, by a small amount.

Then we split the window by time of day. The picture changed.

In the morning, sweeps with a release and sweeps without one did about the same. So in the first hour it is the clock doing the work, not the news. On Fed afternoons the effect is real in the data but rests on a handful of days. What made the filtered result look so good was mostly leaving out the midday sweeps, which lose.

Release spikes run the stops. Liquidity thins just before a release. The first jolt takes price through yesterday's high, where breakout buyers' entries and short sellers' stops sit. Once those orders are filled, no one is left to buy, the late buyers are trapped, and their exits push price back below the level.

Fed afternoons are the clearest case. The 2 pm statement brings a fast first move, and the press conference half an hour later often reverses it as traders re-read the message. Gold reacts more to the Fed than any of our markets: its 15-minute move after the decision is about 5.4 times a normal day's, against about 2.7 times for NQ and ES. That is a bigger stop run to fade, and a sweep of yesterday's high in that window is often the peak of the first reaction.

In the first hour, overnight positions are being adjusted and false breaks are common whether or not there is news. At midday there is no fuel: without news or the opening flow, a poke above yesterday's high has nothing to reverse it, so the fade loses.

You do not need a thousand trades when there is a clear market reason, but the reason tells you where to look, not that it works. The honest next step is to fix the rule before more data comes in, so the coming months can score it without hindsight.

On 8 Nov 2026 we fixed two versions: Fed afternoons only, and the first hour or a Fed afternoon. We will report how both do on the trades that come after this date, including if they fail.

Six Fed days are not enough to tell a real effect from a lucky streak. The event window was chosen after looking at the results, and it fails the honest luck test. Our rebuild and the platform count slightly different event lists, so trade counts differ a little. About 14 months, one market. This is the past, not a promise, and nothing here says how much to trade.

Simulated results on past market data from 2025 and 2026. Research only, not trading signals or advice. Past results do not promise future ones; see the hypothetical performance disclosure at the bottom of the page.