Should you buy gold and hold it, wait for a breakout, or buy a bounce? Holding captures the whole journey, including the falls. Waiting for a signal means spending less time in the market, but risks missing some of the gains.
I compared these three approaches from 2011 to 15 October 2026. The breakout rule buys after gold closes above the highest price traded during the previous 20 trading days. The bounce rule buys after gold touches its 20-day moving average from above and closes back above it. Both use the same stop-loss, profit target and maximum holding period.
Holding made the most money. But buying breakouts captured 78% of that profit while holding a position on just 14% of trading days. Its largest fall from a previous profit peak was about $28,600 per 100 ounces, compared with $159,900 for holding.
The surprise was that breakouts barely improved the odds of gold rising: ten days later, gold was above its signal-day close 55% of the time, against 52% after an ordinary day.
The stronger result came from the trades’ payoffs. That is the comparison that matters here: how much each approach earned, how long it stayed exposed, and how much it lost along the way.
Three approaches to gold
Hold. Buy and keep the position through the rises and falls.
Buy the breakout. Wait for gold to close above the highest price of the previous 20 trading days, then buy at the next daily open. That is the upper line of a 20-day Donchian channel. The trade expects the rise to continue.
Buy the bounce. Wait for gold to approach its 20-day simple moving average from above, touch it and close above it, then buy at the next open. The average uses the previous 20 closes; the trade expects support to hold.
We also test selling short: selling first, aiming to buy back lower. A breakout short follows a close below the previous 20-day low. A bounce short touches the average from below and closes below it.
Both rules use a stop-loss to limit losses and a target to take profits. For buys, the stop is one ATR below entry and the profit target two ATR above. ATR, or average true range, measures recent daily price movement, including gaps. Shorts reverse those distances. If neither price is reached, the trade closes at the end of its tenth trading day.
All results use COMEX gold futures from 2011 through 15 October 2026, after assumed trading and contract-roll costs. Futures prices reflect carrying effects.
A real breakout, and a trade that lost
On 10 January 2025, gold closed at $2,717.40, a dollar above its previous 20-day high. The rule bought at the next open, $2,721, with a stop at $2,688.99 and a target at $2,785.02. Gold hit the stop that day: a loss of about $3,226 on 100 ounces, including trading costs.
The rule worked as written and still lost. What matters is whether the winners pay for the losses.
Holding made the most, and lost the most along the way
For the first comparison, each approach holds the same 100 ounces whenever it has a position. A drawdown is the fall from a previous peak in cumulative profit.
The breakout retained much of the profit with a smaller drawdown
| Per 100oz | Hold | Breakout buys | Bounce buys |
|---|---|---|---|
| Total profit | $191,030 | $149,697 | $24,840 |
| Largest fall from a peak | $159,945 | $28,585 | $48,678 |
| Profit ÷ largest fall | 1.2 | 5.2 | 0.5 |
| Trading days with a position | 100% | 14% | 17% |
The breakout buys gave up about a fifth of holding’s profit. In exchange, they spent 86% of trading days out of the market, and their worst drawdown was less than a fifth of holding’s. Holding paid more, but required sitting through a $159,945 decline from the peak per 100oz.
The bounce offered a weaker balance: much less profit and a larger drawdown than the breakout.
Buying side, every year: profits, average R and entry counts
| Year | Breakout | Bounce | Buy and hold | Buy any day |
|---|---|---|---|---|
| 2011 | $34,470+1.39R · 10 trades | $5,915+0.11R · 10 trades | $13,990 | +0.25R249 overlapping entry dates |
| 2012 | $5,709+0.29R · 8 trades | −$6,431-0.33R · 9 trades | $8,870 | -0.20R250 overlapping entry dates |
| 2013 | −$203-0.01R · 3 trades | −$3,784-0.22R · 8 trades | −$47,890 | -0.28R250 overlapping entry dates |
| 2014 | −$4,439-0.30R · 8 trades | $164-0.02R · 9 trades | −$760 | -0.09R245 overlapping entry dates |
| 2015 | $1,905+0.18R · 6 trades | −$3,166-0.42R · 5 trades | −$12,630 | -0.16R249 overlapping entry dates |
| 2016 | −$6,688-0.35R · 9 trades | $10,786+0.66R · 11 trades | $7,300 | +0.13R249 overlapping entry dates |
| 2017 | −$6,418-0.49R · 10 trades | $7,133+0.59R · 9 trades | $13,180 | +0.19R249 overlapping entry dates |
| 2018 | −$4,985-0.73R · 6 trades | −$5,479-0.33R · 13 trades | −$4,990 | -0.29R250 overlapping entry dates |
| 2019 | $4,902+0.48R · 9 trades | $2,423+0.11R · 8 trades | $19,870 | +0.11R250 overlapping entry dates |
| 2020 | −$1,542+0.40R · 12 trades | $430+0.12R · 9 trades | $31,050 | +0.21R250 overlapping entry dates |
| 2021 | $7,904+0.56R · 6 trades | $5,401+0.18R · 11 trades | −$10,390 | -0.14R248 overlapping entry dates |
| 2022 | $3,440+0.11R · 8 trades | −$970+0.11R · 8 trades | −$4,980 | -0.10R249 overlapping entry dates |
| 2023 | −$7,702-0.19R · 11 trades | −$2,717-0.26R · 8 trades | $12,750 | +0.14R249 overlapping entry dates |
| 2024 | $21,620+0.67R · 13 trades | $20,892+0.66R · 9 trades | $42,780 | +0.40R251 overlapping entry dates |
| 2025 | $86,310+1.17R · 15 trades | −$4,571-0.22R · 10 trades | $152,190 | +0.75R250 overlapping entry dates |
| 2026 to 15 Oct | $15,414+0.50R · 6 trades | −$1,188-0.06R · 7 trades | −$31,435 | -0.07R195 overlapping entry dates |
2026 ends on 15 October. Annual buy-and-hold rows start at that year’s first open and close at its last close, so they are separate annual comparisons and do not sum exactly to one uninterrupted position. Fixed-target trades must have a complete ten-bar window within the year, as in the original test.
The moving-average bounce did no better than an ordinary buy
A rising gold market can reward a buy even without an indicator. To check what the signal added, we also test buying or selling on any eligible day, using the same exits.
1R is the amount initially risked at the stop. An average +0.31R means a gain equal to 31% of that risk per trade.
Same rules: did buying or selling work better?
| Entry | Buy / long | Sell / short |
|---|---|---|
| 20-day breakout | +0.31R140 entries · SE 0.12R | 0.00R105 entries · SE 0.13R |
| 20-day-average bounce | +0.06R144 entries · SE 0.11R | -0.11R124 entries · SE 0.11R |
| Any eligible day | +0.06R3,933 entries · SE 0.05R | -0.05R3,933 entries · SE 0.04R |
The bounce and ordinary buy both averaged about +0.06R. Waiting for the moving average showed no clear improvement. Breakout buys averaged +0.31R.
Gold rewarded buying strength, not selling weakness
The selling results were weaker: approximately zero R per breakout short and −0.11R per bounce short. Markets need not behave alike in both directions. In this gold history, buying strength worked better under these rules than selling weakness.
The breakout also made money below the 200-day average
The 200-day moving average describes the longer-term trend using information available at the time. We group the existing buy signals according to whether gold closed above or below that average on the signal day.
The same buying rules, in both market conditions
| Approach | Above the 200-day average | Below the 200-day average |
|---|---|---|
| Breakout | +0.29R113 entries · standard error 0.14R | +0.30R26 entries · standard error 0.28R |
| Bounce | +0.03R82 entries · standard error 0.14R | +0.09R60 entries · standard error 0.17R |
| Buy any day | +0.11R2,271 entries · standard error 0.06R | -0.02R1,610 entries · standard error 0.07R |
Above the average, buying on an ordinary day returned +0.11R. Below it, that fell to −0.02R, while breakout buys still averaged +0.30R. A short-term breakout can occur below a longer-term average. That below-average breakout result comes from 26 trades.
Selling results above and below the 200-day average
| Short entry | Above the average | Below the average |
|---|---|---|
| Breakout | 0.00R36 entries · SE 0.22R | -0.03R68 entries · SE 0.15R |
| Bounce | -0.21R68 entries · SE 0.14R | -0.01R54 entries · SE 0.18R |
| Any day | -0.09R2,271 entries · SE 0.06R | +0.01R1,610 entries · SE 0.07R |
Same classification and fixed exits as the buying table. Warmup dates remain in the full results. Dollar totals for the separate short streams were $28,238 for breakouts and −$67,289 for bounces. A near-zero average R can coexist with a dollar profit because a fixed 100oz position risks different dollar amounts as ATR changes.
Buy-and-hold in the same conditions, and the 2013–2015 example
A continuous holding has no entry signal or ATR stop. Its daily results can instead be grouped using the previous day’s position relative to the 200-day average, known before that day began.
| Previous close | Average daily return | Days |
|---|---|---|
| Above the average | +0.032% | 2,348 |
| Below the average | +0.012% | 1,692 |
These percentages are daily returns on the gold position, not account returns or R per trade. The 53 warmup days are excluded from this split only. No buying or selling filter is applied.
As a separate historical example, buying and holding 100oz lost $62,890 in 2013–2015. The original breakout lost $2,736 and the bounce lost $6,786. The breakout reduced the loss largely by staying out of the market. Those dates describe a known decline; they were not used to select trades.
The payoff mattered more than predicting direction
Ten trading days after an upward breakout, gold was above the signal close in 55% of cases. After an ordinary day: 52%. The breakout barely improved that probability. Its stronger average trade came from the size of gains and losses under the tested entry and exit rules.
Changing the exits also changed the outcome. A second breakout version used a wider two-ATR stop, no profit target and no time limit. It sold a long after a close below the previous ten-day low, at the next open. Shorts used the opposite rule.
That version earned $129,329 from buys, versus $149,697 with the fixed target, and held a position on 37% of days rather than 14%. Letting winners run did not automatically improve the result.
The channel-exit breakout also beat 182 of 200 random-entry portfolios over the full period. It earned $129,329 against their median of $52,434, using the same exits and approximately matched time in the market. That supports the value of its entry timing in this run. In the separate 2016–2020 and 2021–2025 tests, however, it finished near the random median.
The two exit versions side by side
Two exit rules, applied to the same breakout idea
| Fixed target | Channel exit | |
|---|---|---|
| Initial stop | 1 ATR below entry | 2 ATR below entry |
| Take profit / leave | 2 ATR above entry | Close below the previous ten-day low; sell at the next open |
| Time limit | Ten trading days | None |
| Average result | +0.31R | +0.40R |
| Profit per 100oz | $149,697 | $129,329 |
| Days holding a position | 14% | 37% |
| Comparison without a signal | Buy-any-day: +0.06R per trade | Random-long median: $52,434; middle 80%: −$56,948 to $125,485 |
Explore the alternative exits and random-entry comparison
Channel exits: longs, shorts and random-long controls
| Period | Breakout long | Breakout short | Random long: median and range |
|---|---|---|---|
| 2011–2015 | $13,41928.2% exposed | $11,845 | −$16,77310th–90th: −$48,595 to $16,08628.3% mean exposure |
| 2013–2015 | −$25,40623.3% exposed | $10,990 | −$16,78510th–90th: −$30,950 to −$2,91822.9% mean exposure |
| 2016–2020 | $26,42742.3% exposed | −$42,752 | $27,26010th–90th: −$3,326 to $45,42542.2% mean exposure |
| 2021–2025 | $39,99139.9% exposed | −$78,934 | $39,48610th–90th: −$7,474 to $97,31741.9% mean exposure |
| 2026 to 15 Oct | −$45,16217.6% exposed | $35,830 | −$23,65710th–90th: −$60,068 to $16,79015.8% mean exposure |
| 2011–2025 | $78,09936.8% exposed | −$113,010 | $39,32610th–90th: −$33,124 to $103,18137.1% mean exposure |
| 2011–15 Oct 2026 | $129,32936.8% exposed | −$77,180 | $52,43410th–90th: −$56,948 to $125,48536.5% mean exposure |
The random comparison is a research check, not a suggestion to trade randomly. Separate period rows restart the test, so they cannot be added to recreate one continuous run. Long and short results are separate streams. The short rule helped in the bear years but lost heavily after 2016.
What to take from it
The 20-day Donchian breakout was a useful upside entry signal in this gold history. With the tested stop, target and time limit, it produced a better average payoff than the moving-average bounce or buying on an ordinary day.
Its attraction was the balance: 78% of holding’s profit, a position open on 14% of trading days, and a much smaller drawdown. A good strategy does not have to capture every dollar of a rise. Spending less time exposed is valuable when it preserves enough profit and reduces the losses along the way.
Other indicators and exits can change the result. This comparison gives us a starting point: judge the profit captured, the time exposed and the depth of the losses together.
Glossary
Open a term for its meaning in this article.
Long / short
A long position gains when gold rises. A short gains when gold falls: sell first, then buy back. Either can lose when the price moves the other way.
Breakout / Donchian channel
The channel marks the highest and lowest prices of the preceding 20 trading days. This test buys after a daily close above the upper line, or sells after a close below the lower line. Entry is at the next daily open.
Moving average / bounce
A simple moving average is the average closing price over a chosen number of days. Here the bounce uses the previous 20 closes: buy after a touch from above and a close above the line; sell after a touch from below and a close below. The separate 200-day average describes the longer-term trend.
ATR: average true range
A measure of recent daily price movement, including gaps. This test uses a smoothed 14-day ATR, known before the signal day, to set the stop and target distances. It measures movement, not direction.
Stop-loss / profit target / channel exit
A stop exits a losing trade; a target exits at a planned profit. The fixed-target version uses a one-ATR stop, two-ATR target and ten-day limit. The alternative uses a two-ATR stop and a close through the opposite ten-day channel, with no target or time limit. Channel exits execute at the next open.
R / average result
One R is the initial distance from entry to stop, multiplied by position size. An average +0.31R means a gain equal to 31% of initial risk per trade on average. It is not an account return. Each trade uses its own risk amount; the two exit versions therefore use different R units.
Standard error / uncertainty
Standard error describes uncertainty around an estimated average. It does not describe the range of individual trade outcomes. Here it allows for related nearby observations, but not for selecting rules after testing many alternatives.
Win rate / payoff
Win rate is the share of trades with a net profit. Payoff describes how large wins and losses are. A strategy can win less than half the time and still make money if its gains are large enough.
Drawdown / exposure
Drawdown is the fall from a previous peak in cumulative profit. This study measures it at daily closes, so intraday falls can be larger. Exposure is the share of trading days with a position open, not the number of broker funding charges.
Any-day / random-entry comparison
These check what happened without an indicator signal, using the same exits. Any-day outcomes overlap and are not one portfolio. The channel test instead uses 200 random-entry portfolios with approximately matched time in the market.
Costs / funding / futures roll
The test deducts $25 per 100oz round trip and additional roll costs when replacing an expiring contract. Futures prices reflect carrying effects. Historical CFD holding charges are not deducted: these are futures results, not broker-account returns.
Follow-through
Further movement after a signal. The final table checks the closing price five and ten trading days later against the signal-day close. It does not apply a stop or target and is not a trade win rate.
Test period / partial year
2011 through 15 October 2026. The partial 2026 results are not annualised. Fixed-target entries need a complete ten-day window within the calendar year; channel trades can cross year-end. This is historical research, not untouched validation data.