Rules for a Gold Rush
How the Market Wave Algorithm for ETFs beat monthly dollar-cost averaging in every GLD window we tested, from a record bull run to a worst-case 2011 entry it turned into $50,398 of banked cash.
The same Market Wave algorithm our crypto and equities clients run every day, applied unchanged to GLD, the largest physically backed gold ETF in the world.
The same algorithm, pointed at the hardest asset to hold with a clear head
Our first ETF study took on QQQ and the index-fund playbook on the friendliest terrain passive investing has. This study points the same algorithm at a very different kind of asset.
Gold is the oldest store of value in the world, and the hardest one to hold with a clear head. It pays no dividend and produces no earnings, so there is no valuation anchor to steady your judgment. It spends years asleep, then reprices violently in months. Every investor who has owned it knows both feelings: the fear of buying too late, and the regret of never selling.
As we write this in July 2026, gold sits roughly 28% below the all-time high it set in late January. Both feelings are fresh.
We did not build a new model. We took the exact Market Wave algorithm our clients already run and pointed it at GLD. The benchmark in every test is a monthly dollar-cost-averaging investor: the same capital, split into equal monthly buys, never selling, ignoring price.
The record bull run
Strategy+55.5%
- DCA
- +22.5%
The mid-trend entry
Strategy+20.7%
- DCA
- +2.7%
The worst case
Strategy, 100% cash+50.4%
- DCA
- +40.3%
All three runs started from the same hypothetical $100,000 in cash and zero shares.
Backtested performance. Past performance does not guarantee future results. For educational purposes only.
The metal that tests every investor
GLD is the largest physically backed gold ETF in the world. When you buy it, you buy gold's entire emotional weather system: inflation fear, currency politics, war premiums, and the oldest safe-haven instinct in finance.
Gold's price, at the levels that matter:
- Sep 2011$1,920 peak
- Dec 2015$1,050 low
- Aug 2020$2,075
- Mar 2024$2,180
- Jan 2026$5,589 all-time high
- Jul 2026About $4,060, down 28%
GLD shares
| Date | Aug 2011 | 2015 low | Jun 2024 | Jan 2026 | Jul 2026 |
|---|---|---|---|---|---|
| GLD share price | ~$158 | ~$100 | $211.60 | ~$500 | $370.80 |
| Scenario window | Scenario 3 opens | Scenario 1 opens | Scenarios 1 and 2 end |
Scenario 3, worst case: weekly, Aug 2011 to Mar 2024. Scenario 1, record bull run: weekly, Jun 2024 to Jul 2026. Scenario 2, mid-trend entry: daily, Apr 2025 to Jul 2026. Price levels per public market data and World Gold Council figures. For educational purposes only. Not a recommendation to buy or sell any security.
Gold rewards patience enormously, but only across cycles that outlast most investors' patience, and it punishes bad entry timing for years at a stretch. It is precisely the kind of asset where emotion does the most damage, in both directions.
Too early, too late, or asleep
Every gold investor lives somewhere in one of three mistakes.
Too early: buying the fever, holding the fever
The investor who bought the 2011 top on inflation headlines watched 45% of their capital erode over four years, then waited nearly nine years just to get back to even. Gold did eventually vindicate them, but most did not stay for it.
Hope is not a position-management plan.
Too late: deciding you missed it
When an asset has already doubled, the reflex is paralysis. The investors who force themselves in late tend to do it emotionally, all at once, near a top, then hold through the reversal because selling would mean admitting the mistake.
Gold's 28% correction from the January 2026 peak is this story playing out in real time.
Asleep: never taking a profit
Gold produces no dividend and no earnings. Until you sell some, a gold position is a number on a screen. Buy-and-never-sell works passably for an index fund with earnings behind it.
On gold it leaves the investor with nothing banked after years of being right.
Dollar-cost averaging is the standard prescription, and it deserves respect: it removes the emotion, spreads the entry, and beat doing nothing in all three of our windows.
But it hard-codes the third mistake forever. By design it does not look, and it never, ever sells.
What Market Wave actually does
Arch Public's trend and momentum capture algorithm, the same algorithm already running for our crypto and equities clients, applied to GLD unchanged. It does three things with rules instead of emotions.
Ride confirmed directional moves
It enters on rule-based confirmation of a trend, not on a headline, a fear spike, or a fixed calendar date.
Exit on rules, not nerves
It takes profits into strength on defined triggers, banking gains into cash rather than holding every position on principle.
Step aside cleanly when the trend is not there
In chop, indecision, and multi-year declines it is willing to hold cash and wait, exactly where a calendar-driven buyer keeps buying anyway.
The only variables in this study are the time window and the chart interval. Same algorithm, same starting $100,000.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
“The record bull run”
June 1, 2024 to July 8, 2026 · Weekly chart
The gift window, and the one where discipline is hardest to keep. Everyone made money here. The question is how much, and how.
GLD opens near $212 and rides gold's historic advance through $3,000, $4,000, and the January 2026 record, then absorbs the 28% correction.
A rising market does not mean there was nothing to do. The algorithm treated every overextension as an opportunity, selling into strength and buying back lower, banking $43,105 in completed-trade profit while still building a 411-share position. One method only accumulated; the other accumulated, harvested, and repositioned.
+55.5%
- DCA +22.5%, outperformance +33.1 percentage points
- Strategy ending value
- $155,517
- Monthly DCA ending value
- $122,462
- Realized profit, banked
- $43,105
- Unrealized, open
- +$12,412
- Total profit and loss
- +$55,517
- Position held
- 411 at $341 avg cost
Open full-size chart
Backtested from a weekly-chart trade schedule, valued July 8, 2026. Past performance does not guarantee future results.
“The mid-trend entry”
April 1, 2025 to July 8, 2026 · Daily chart
The too-late test. This window opens with gold already above $3,000. This is the exact moment most investors freeze: chase it or skip it.
Gold peaked in late January 2026 and fell roughly 28% by early July.
The algorithm had already done the thing the late human buyer never does: it sold into the top it did not know was the top, banking $33,377 in realized profit on the way up. The mid-trend entry everyone fears was worth 20.7% to the rules, and 2.7% to the calendar, over the identical fifteen months.
+20.7%
- DCA +2.7%, outperformance +18 percentage points
- Strategy ending value
- $120,700
- Monthly DCA ending value
- $102,732
- Realized profit, banked as cash
- $33,377
- Total profit and loss
- +$20,700
- Position held
- 326 shares at $409 avg
Open full-size chart
Backtested from a daily-chart trade schedule, valued July 8, 2026. Past performance does not guarantee future results.
“The worst case”
August 1, 2011 to March 11, 2024 · Weekly chart
The window we would hide if this were marketing: a 45% decline over four years, then most of a decade grinding back. An investor who bought the open waited roughly nine years to break even.
The window opens five weeks before gold's September 2011 peak. GLD did not durably reclaim its 2011 high until 2020. Across nearly thirteen years the algorithm sold rallies, stepped aside, and re-entered on confirmation.
Every dollar of the DCA gain is unrealized, sitting in 640 shares never once trimmed in 151 months. One investor spent thirteen years accumulating a paper gain and still owns all of the next drawdown. The other finished holding money.
+50.4%
- DCA +40.3%, outperformance +10.1 percentage points
- Strategy ending value
- $150,398
- Monthly DCA ending value
- $140,331
- Total bought, closed entries
- $112,633.49
- Total sold, closed exits
- $163,031.68
- Realized profit, all cash
- $50,398.19
Backtested from a weekly-chart trade schedule, valued March 11, 2024. Past performance does not guarantee future results.
Three windows, one pattern
Every test measured against the same disciplined monthly DCA benchmark. In every window the algorithm finished ahead, and finished with meaningful realized cash, which DCA by definition never produces.
| Scenario | The record bull run | The mid-trend entry | The worst case |
|---|---|---|---|
| Chart, window | Weekly · 2024 to 2026 | Daily · 2025 to 2026 | Weekly · 2011 to 2024 |
| Ending value | $155,517 | $120,700 | $150,398 |
| Strategy ROI | +55.5% | +20.7% | +50.4% |
| DCA ROI | +22.5% | +2.7% | +40.3% |
| Vs DCA | +33.1 | +18.0 | +10.1 |
| Realized profit, banked as cash | $43,105 | $33,377 | $50,398 |
| DCA realized profit | $0 | $0 | $0 |
Every window started from the same $100,000 in cash and zero shares. Realized profit is completed-trade profit per window. This is what DCA can never produce.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
Gold only pays the seller
The instinct is to assume the algorithm predicted gold's moves. It predicted nothing. It responded to structure, and on this asset, responding is worth more than anywhere else.
A stock or an index fund has earnings compounding underneath it. Gold has no earnings and pays no dividend. One hundred percent of a gold investor's return is price change, which means one hundred percent of the return is only ever collected by selling.
A method that never sells has locked itself out of the only payout window the asset has.
In the bull run: $43,105
DCA kept a fine paper gain. Only the algorithm converted the overextensions into banked profit while still riding the trend.
In the mid-trend entry: $33,377
DCA's calendar marched into the top at full size; mean reversion took it back to 2.7%. The algorithm sold strength it did not know was the top.
In the worst case: $50,398
Thirteen years of DCA discipline produced $0 realized. The same thirteen years of rules produced cash and zero exposure left on the table.
Dollar-cost averaging, realized profit across all three windows: $0.
DCA removes emotion by removing sight. On gold, sight is the whole game, because the asset never writes you a check. The algorithm keeps the discipline and keeps its eyes open.
For educational purposes only. Not investment advice.
The calendar is not a strategy
- 01
The DCA prescription gets the diagnosis right
Emotion is the number one destroyer of gold returns: panic buying at fever tops, capitulating at bottoms, freezing when the trend leaves. DCA cures the emotion and beat doing nothing in all three windows. The diagnosis deserves real respect.
- 02
But the cure amputates the payout
To guarantee it never panics, DCA agreed to never look and never sell. On an asset whose entire return is price, never selling means never collecting. Across all three windows, disciplined DCA realized a combined $0. Real gains, but all story, no settlement.
- 03
Discipline and awareness are not opposites
The same rules that remove emotion can watch price structure, take profits into strength, step aside in declines, and re-enter on confirmation. That combination beat the calendar in every window, and it is the only approach that ever put realized cash in the account.
Gold will keep doing what it has done for fifteen years: long sleeps, violent repricings, euphoric tops, punishing corrections. You do not have to predict any of it. You have to be systematically present for it, and systematically willing to get paid.
For educational purposes only. Trading ETFs involves significant risk.
Rules, not reactions
Arch Public's agentic strategies now run on gold through GLD, the same rules-based Market Wave algorithm our crypto, equities, and index ETF clients already use, applied unchanged.
Zero custody. Your assets never leave your account.
Full control. Pause, adjust, or stop any strategy at any time.
Rules, not reactions. Neither the fear of being late nor the refusal to sell ever makes the call.
How much of your capital is invested with its eyes closed, and how much of what it earns will you ever actually collect?
Across a euphoric run, a late entry into a top, and a thirteen-year worst case, the algorithm banked between $33,377 and $50,398 in completed-trade cash per window. The dollar-cost averager, across all three combined, realized nothing. On an asset with no dividend and no earnings, that is the whole difference between owning gold's story and getting paid by it.