The short answer
We tested the two classic ways to trade yesterday’s gold high and low on two years of hourly data (513 CME trading days, October 2024 to October 2026):
- Break: buy the first trade above the previous day’s high (or sell below the low), stop at the previous day’s midpoint, target 1R, flat at the end of the session. 439 trades, 51.5% winners, +$308 per ounce after a $0.30 cost. Profit factor 1.05: positive, but thin, and almost all of it came in the first year.
- Fade: sell at the previous day’s high (or buy at the low), stop a quarter of the prior range beyond it, target the prior midpoint. 435 trades, 35.6% winners, −$665 per ounce.
On daily bars since 2015, a close-based version of both lost. The previous day’s high and low are useful reference levels; on their own, neither the break nor the fade was a dependable edge.
The levels and how often they are hit
The previous day’s high (PDH) and low (PDL) are where yesterday’s buyers and sellers gave up. Stops and breakout orders tend to sit just beyond them, which is why they attract price. Before testing any rule, we counted how often gold reached them.
| Outcome on the next trading day (513 days) | Days | Share |
|---|---|---|
| Traded above PDH only | 220 | 42.9% |
| Traded below PDL only | 180 | 35.1% |
| Traded through both | 48 | 9.4% |
| Stayed inside the previous range | 65 | 12.7% |
About 87% of days took out at least one level, so “price will visit yesterday’s high or low” is close to a given. The real question is what happens next.
Exact rules
Version A: the breakout
| Rule | Detail |
|---|---|
| Trading day | CME day: 18:00 to 17:00 New York time (US daylight saving handled for every date) |
| Orders | Buy stop at PDH and sell stop at PDL, placed before the session opens; the first one filled cancels the other |
| Stop | Previous day’s midpoint (half the prior range away; about $41 on average) |
| Target | 1R (a further half of the prior range beyond the level) |
| Exit | Close at the end of the session if neither is hit |
Version B: the fade
| Rule | Detail |
|---|---|
| Orders | Sell limit at PDH and buy limit at PDL; first fill cancels the other |
| Stop | A quarter of the prior day’s range beyond the level |
| Target | The prior day’s midpoint (2R) |
| Exit | End of the session |
Version C and D: daily-close versions since 2015
On daily bars: if the day closes above PDH, buy the next open and hold one day (sell for a close below PDL). Version D does the opposite (fade the close).
How we tested it
Data. COMEX gold futures (GC=F) 1-hour bars from 30 October 2024 to 30 October 2026 (11,445 bars), grouped into CME trading days in New York time, and daily bars from January 2015. Execution. Resting orders fill at the level or at the bar’s open if price gaps through it. Inside the fill bar we assume the usual price path (an up bar goes open, low, high, close; a down bar goes open, high, low, close) and only count stops or targets reached after the fill; if both could be hit, the stop counts. Limit fills never count a target on the fill bar. Cost. $0.30 per ounce per round trip. Results in US dollars per ounce, 1 ounce per trade.
The results
Hourly, two years (513 trading days)
| Version | Trades | Win rate | Avg per trade | Profit factor | Net | Max drawdown | Year 1 | Year 2 |
|---|---|---|---|---|---|---|---|---|
| A: break, stop at midpoint, 1R | 439 | 51.5% | +$0.70 | 1.05 | +$308.20 | $958.40 | +$454.90 | −$146.70 |
| B: fade, target midpoint | 435 | 35.6% | −$1.53 | 0.88 | −$665.40 | $1,412.30 | −$187.00 | −$478.40 |
Year 1 runs to 30 October 2025. The breakout made $774 on longs and lost $466 on shorts.
Daily closes, since 2015
| Version | Trades | Win rate | Profit factor | Net | 2015–2020 | 2021–2026 |
|---|---|---|---|---|---|---|
| C: close beyond PDH/PDL, follow for 1 day | 1,569 | 44.9% | 0.91 | −$1,040.70 | +$117.90 | −$1,158.50 |
| D: close beyond PDH/PDL, fade for 1 day | 1,567 | 47.1% | 0.97 | −$332.70 | −$563.30 | +$230.70 |
Costs (version A)
| Cost per round trip | Net |
|---|---|
| $0 | +$439.90 |
| $0.30 | +$308.20 |
| $0.60 | +$176.50 |
What the results say, in plain English
- The breakout was slightly better than a coin flip. A profit factor of 1.05 means it made $1.05 for every $1 lost. Over 439 trades that is $308 an ounce, with a drawdown three times as large ($958).
- It worked in year 1 and not in year 2. Year 1 (+$455) was a steady bull trend; year 2 (−$147) was a violent swing from about $4,000 to above $5,600 and back. In choppy conditions, yesterday’s high is broken and then reclaimed.
- The fade lost in both years. Selling the first touch of yesterday’s high with a tight stop was stopped out too often, with a worst run of 17 losers in a row. Gold tends to push through the level by more than a quarter of the prior range before any rejection.
- Daily closes were noise. Following or fading a daily close beyond PDH or PDL lost over eleven years and flipped sign between periods.
- Costs did not decide the result. Doubling the cost to $0.60 kept version A positive at $177, because the trades are large relative to the cost.
When the previous day high/low strategy fails
- Both levels on one day. On 48 of 513 days gold went through both. A breakout trader takes a loss on the first break and may miss the real move.
- News days. CPI, payrolls and Fed days often spike through a level and reverse within minutes.
- Inside days. On 65 days neither level was reached, so there was no trade. That is fine, but it means a breakout trader must accept quiet days.
- Wide prior ranges. After a huge day, the midpoint stop becomes enormous. Size down or skip.
How traders add context
The numbers suggest PDH and PDL are better as places to watch than as automatic triggers. Common refinements:
- Sweep or hold. Wait for an hourly or 15-minute close. A close back inside the range after taking the level is a sweep; a close and hold beyond it is a breakout. Our liquidity sweep lesson teaches this distinction.
- Session timing. Breaks in the New York morning behave differently from breaks in the quiet Asian session.
- Confluence. A previous day high that lines up with a weekly level or a daily pivot carries more weight. See support and resistance on gold and pivot points for gold.
Risk sizing
The breakout stop averaged about $41 per ounce, because the prior day’s range was often $80 or more. Risking $100:
- $100 / ($41.23 + $0.30) is about 2.4 ounces (0.02 lots).
- At that size, the $958 drawdown per ounce would have been about $2,300.
Size every trade from the stop with the position size calculator, and expect stop distances to change daily with volatility.
Checklist before you trade yesterday’s levels
- Use one day boundary and stick to it: the CME day (18:00 New York) or your broker’s server day. Mixing them moves every level.
- Mark the midpoint too. It was our breakout stop and our fade target, and it is often where a failed break turns.
- Note the prior range. After a very wide day the midpoint stop is huge; size down or wait for a tighter structure.
- Wait for a close. A touch is not a break. Decide whether you need a 15-minute or hourly close beyond the level before acting.
- Check the calendar. On CPI, payroll and Fed days, both levels are more likely to be hit.
- Journal every touch as a break, a sweep or nothing, so that your own sample builds up alongside our numbers.
Test it yourself
Mark yesterday’s high, low and midpoint on your chart every morning for a month and record, without trading, what happens at each touch: break and hold, sweep and reverse, or nothing. Then test the rules above on your platform’s data, using your broker’s day boundary. The strategies tested hub compares this with 13 other named strategies.
Practise it with other traders
Previous-day levels are simple to mark and hard to trade well, which makes them perfect for shared practice. In the XAUUSD Global community, traders post their levels each morning and review the touches together at the close. The first month is free, then $19.99 a month.
Educational content only, not financial advice. Trading gold with leverage carries a high risk of loss.
Frequently asked questions
How often does gold break the previous day's high or low?
On 513 CME trading days from October 2024 to October 2026, gold traded above the previous day’s high only on 220 days, below the low only on 180 days, through both on 48 days, and stayed inside on 65 days. So about 87% of days took out at least one of the two levels.
Is it better to trade the break or the fade of the previous day's high?
In our two-year hourly test, the breakout version (stop at the previous day’s midpoint, 1R target) made $308 per ounce over 439 trades after costs, a thin profit factor of 1.05. The fade version (sell at the high, buy at the low) lost $665 per ounce over 435 trades.
Does the previous day high/low strategy work on daily charts?
As a daily-close rule, no. Buying a close above the prior day’s high and selling a close below its low, held for one day, lost $1,041 per ounce over 1,569 trades since 2015. Fading those closes lost $333.
What time does the gold trading day start for previous-day levels?
We used the CME futures trading day, which starts at 18:00 New York time and ends at 17:00 the next day. Many traders instead use their broker’s server day (often 00:00 server time). The levels differ slightly, so be consistent.
Where should the stop go on a previous-day-high breakout?
Our tested version used the previous day’s midpoint, which gave an average stop of about $41 per ounce over the test. That is wide; many traders use a tighter structural stop and accept more stop-outs.
Is a previous day high sweep the same as a breakout?
No. A sweep trades above the high and quickly closes back below it; a breakout holds above it. Our liquidity sweep lesson covers how to tell them apart in real time.
Sources and further reading
- CME Group — Gold futures overview — Contract specifications and trading hours for COMEX gold
- CFTC — Commitments of Traders — Weekly futures positioning, including gold
- Investor.gov — Stop order — How stop orders work, and their limits in fast markets
External sites are independent of XAUUSD Global. Check dates and data yourself.