Gold Scalping Session by Session: The Busiest Hours Paid the Least on 13,736 XAUUSD Bars

Almost every gold scalping guide gives the same timing advice, and my own gold scalping indicator guide gave it too: trade the London open and the London–New York overlap, when liquidity is deepest and gold moves the most. It sounds obviously right. More movement should mean more opportunity.

My gold indicator page promised to leave gold-only claims until the gold bars had been through the same code as EURUSD. This is that run. Same method as every other test on this site: fixed rules, every signal counted, nothing picked by eye.

What data and rules were used?

The data is 13,736 5-minute bars of COMEX gold futures (symbol GC=F) from June 16 to August 26, 2026. Futures trade slightly above spot XAUUSD, but the bar-to-bar movement is close enough that scalping rules behave the same way. Over the period gold went from $4,350.80 to $4,674.90, with a low of $3,955.40 and a high of $4,755.00.

All times are UTC. The four sessions are Asia (22:00–07:00), London (07:00–12:00), the London–New York overlap (12:00–16:00) and the New York afternoon (16:00–21:00). A trade belongs to the session in which it was opened.

The four rules are the ones gold scalpers actually use:

• EMA 9/21 cross — buy when the 9 EMA crosses above the 21, sell on the opposite cross. Why exponential rather than simple averages is covered in my EMA vs SMA test.
• RSI 7 fade 20/80 — buy when RSI 7 drops below 20, sell when it rises above 80.
• 1-hour breakout — buy when a bar closes above the highest high of the previous 12 bars, sell on a close below the lowest low.
• Bollinger 20/2 fade — buy on a close below the lower band, sell on a close above the upper band. The EURUSD version of band trading is in my Bollinger squeeze test.

Each trade enters at the close of the signal bar and exits 6 bars (30 minutes) or 12 bars (60 minutes) later. Trades never overlap and never run across a weekend or the daily futures break. Every trade pays $0.30 per ounce — an assumed round-trip cost for a retail gold account, not a measured one, so every rule is also shown at $0.60 further down.

Which gold session moves the most?

The conventional wisdom is right about movement. The overlap holds 17.2% of the bars but 25.4% of all close-to-close movement, and its median 5-minute bar is 60% bigger than the Asian one. By single hour, the busiest was 13:00 UTC with a median bar of $7.50, and the quietest was 20:00 UTC at $2.30.

Two things the guides skip. Asia is not dead on gold: its median bar of $4.00 is bigger than London’s $3.80, and the 01:00 UTC hour ran at $6.20 — only 13:00 and 14:00 UTC had bigger bars. And the New York afternoon, the session most guides tell you to “tighten targets” in, is the quietest of the four.

Which session actually paid a scalper?

Here is every rule in every session. Each cell is the average result per trade in dollars per ounce after the $0.30 cost, with the number of trades in brackets.

Read it column by column and the pattern is hard to miss:

The session with the most movement came last. The session with the least came first. Every cell has at least 42 trades, so none of this rests on a handful of lucky signals — though the best single cell, RSI 7 fade held 60 minutes in the New York afternoon at +$3.18, is also the thinnest at 42 trades, and I would not build a system on it alone.

Why did the overlap lose money?

Look at the breakout row. A close above the last hour’s high in the overlap is exactly the setup the “trade the overlap” advice is built for — and it lost $1.56 per trade at 30 minutes and $1.17 at 60 minutes. In the New York afternoon the same rule made money at both holds.

Big bars cut both ways. The overlap is where US data lands, and a $6.40 median bar means price regularly travels past a 1-hour high and comes back inside the same half hour. A breakout rule buys that spike at its close and holds it through the reversal. My news release test on EURUSD saw the same thing from another angle: the release candle was 2.48 times a normal bar, yet its direction matched the next 30 minutes only 49.2% of the time. Size is not direction.

The quiet afternoon rewards the opposite behaviour. Moves are smaller, but they are more orderly, and the two fade rules — RSI 7 and Bollinger — made money there at both holds. Movement is not opportunity. Movement that reverses before your exit is a cost.

Is the result just gold going up?

Partly, and it needs saying plainly. Gold gained $324 over the test, and in every one of the eight rules the long trades did better than the short ones:

That drift lifts every session equally, though, so it cannot explain why the same rules did better in the afternoon than in the overlap. It does mean the absolute numbers flatter gold. In a falling or sideways month, expect the positive cells to shrink.

Do the rules survive a higher spread?

Without any session filter, four of eight rules are positive at $0.30 and only two stay positive at $0.60. Both survivors hold for 60 minutes, and three of the four 60-minute rows beat their own 30-minute versions — only the Bollinger fade got worse. On a 30-minute hold the cost is a bigger share of the move, and it shows. If you scalp with RSI on EURUSD as well, the period question is covered in my RSI 7 vs RSI 14 scalping test.

What I would do with this as a gold scalper

• Stop assuming the overlap is the best window. It had the biggest bars and the worst results: 2 of 8 tests positive, -$0.41 per trade on average.
• Test the New York afternoon first. 7 of 8 tests were positive there. It is a quiet session, so your targets have to be realistic for a $3.40 median bar.
• Do not trade breakouts into the overlap. The 1-hour breakout lost in the overlap at both holds, over 270 trades combined.
• Hold for 60 minutes, not 30. Three of the four rules did better at the longer hold, and the two that survive a $0.60 spread are both 60-minute rules.
• Know your real spread. The difference between $0.30 and $0.60 turned four positive rules into two. The method for measuring your own is in my spread and swap cost test.
• Treat the long bias as luck until proven otherwise. Gold rose $324 during the test. A rule that only made money on the buy side is a rule that rode the trend.

The session idea itself is not new on this site: my session vs all-day trading test and the best hours to trade EURUSD both found that when you trade moves results more than which indicator you use. Gold confirms it — just not in the session everyone points to.

Frequently asked questions

What is the best time to scalp gold?

In this test, the New York afternoon, 16:00–21:00 UTC. Seven of eight rule tests were positive there after a $0.30 cost, averaging +$0.85 per trade. The London–New York overlap, the usual recommendation, averaged -$0.41 with only two positive tests.

Is the London–New York overlap good for gold scalping?

It is the most volatile session, with a median 5-minute range of $6.40 and 25.4% of all movement in 17.2% of the bars. But the scalping rules lost there on average, and the 1-hour breakout lost $1.56 per trade over 155 trades. Volatility that reverses within your hold time works against you.

Is the Asian session bad for trading XAUUSD?

Not because it is quiet — gold’s median Asian bar of $4.00 was larger than London’s $3.80. Results there were close to flat: 2 of 8 tests positive and -$0.08 per trade on average. It was neither the best nor the worst session.

Which scalping indicator worked best on gold?

Across all sessions, the EMA 9/21 cross held for 60 minutes: 366 trades, 52.7% wins and +$0.56 per trade after a $0.30 cost, still positive at $0.60. The best single session result was RSI 7 fade in the New York afternoon at +$3.18, but on only 42 trades. Settings for these rules are compared in my gold indicator guide.

Does this data use XAUUSD or gold futures?

COMEX gold futures, symbol GC=F. Futures sit slightly above spot XAUUSD but move almost bar for bar with it, so session behaviour and rule results carry over. Your broker’s price will differ slightly, and your spread will decide whether the small positive rules stay positive.

Why use dollars per ounce instead of pips?

Brokers define a gold pip differently — some at $0.01, some at $0.10 — so a pip number on gold means different things on different accounts. Dollars per ounce is the same everywhere. To convert, multiply by your contract size: a standard 100-ounce lot turns +$0.56 per ounce into +$56 per trade.