Skip to the case study
All case studies
Contents
The results
The subject: Advanced Micro Devices
The investor's dilemma
The algorithm
“The all time high scalp”
“The cool off rehearsal”
“The getting paid year”
Side by side
Why the results differ
The real lesson
A path forward
Book a demo
Case studies
AMD
Case study: Arch Public for stocks
AMD: the stock that never told the same story twice
The Market Wave Algorithm for Equities against AMD in three windows. Buy and hold beaten in all three. Dollar cost averaging beaten in all three.
Advanced Micro Devices: the chipmaker that went from AI darling to tariff casualty to the strongest large cap story of 2026, all inside thirty months.
AMD
Stock
Market Wave Algorithm for Equities
Valued August 12, 2026
10 min read
Why this case study, and why now
We have already shown what disciplined, rules-based execution does with a runaway winner, a brutal drawdown, a sideways index, and a stock everyone had already given up on. AMD is a different problem. AMD kept changing its story, and every new story asked investors to reverse their last decision.
From January 2024 through today, AMD holders were asked to believe six different things in sequence: that AMD was the next great AI trade, that it had lost the AI race, that tariffs had broken the semiconductor cycle, that a single OpenAI deal changed everything, that 2026 was a once-in-a-decade run, and that the run was over. Each belief arrived with a violent price move attached.
That history leaves today's investor with three questions, and this study answers each one with its own window and its own recipe of the Market Wave Algorithm for Equities.
Is it too late to buy a stock at all time highs?
The micro recipe scalps AMD hour by hour through the record close and the pullback after it. May 1, 2026 to August 12, 2026, hourly chart, micro scope.
Strategy
+44.3%
Buy and hold
+38.3%
DCA
+7%
What if the AI trade cools off and markets trend down for a year or two?
The round trip runs the algorithm through AMD's last full boom, bust, and recovery. January 1, 2024 to January 1, 2026, daily chart, mid scope.
Strategy
+57.9%
Buy and hold
+54.5%
DCA
+49.2%
What did this year actually pay?
The macro recipe is the 2026 scoreboard. January 1, 2026 to August 12, 2026, four hour chart, macro scope.
Strategy
+145.3%
Buy and hold
+120.2%
DCA
+68.9%
Every run starts from the same hypothetical $100,000 in cash and zero shares.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
Advanced Micro Devices, AMD
The subject
Advanced Micro Devices designs the CPUs and GPUs inside data centers, PCs, and game consoles, and is the only company mounting a serious challenge to Nvidia in AI accelerators. Its data center business is now the center of the story: AI chips, and the racks they live in.
The price history reads like four different stocks.
January 2, 2024 AMD opens the year near $139, riding the first AI wave.
March 2024 The wave crests. AMD closes above $211 at its then all time high, up roughly 50% in nine weeks.
March 2024 to April 2025 The long unwind. AI accelerator revenue disappoints against Nvidia, then the April 2025 tariff shock hits semiconductors hardest. AMD touches a two year low near $76 on April 8, 2025, roughly 65% below the 2024 peak.
October 2025 The comeback. A multi gigawatt OpenAI partnership sends the stock up 24% in a day, to a late October close near $263.
January 2026 AMD enters the year near $214, still below the October high.
First half of 2026 The supercycle year. Blowout first quarter earnings, an agentic AI CPU demand cycle, and a 6 gigawatt Meta deployment deal drive AMD to a record close of $580.91 on June 30, 2026.
July to August 2026 The stock tumbles 18% in July on profit taking, beats second quarter estimates in early August, and sells off anyway. It trades near $487 today.
Here is the catch that matters for this study. Every one of those chapters produced a real gain for somebody. Almost nobody captured more than one of them, because each chapter began by punishing whoever believed the last one.
The investor's dilemma
AMD is a machine for manufacturing bad decisions. Not because holders are foolish, but because the stock keeps presenting setups that human psychology handles badly.
Narrative whiplash
Six stories in thirty months, each arriving with a headline and a gap. The investor who repositioned on every story bought the March 2024 top, sold the April 2025 bottom, and watched the OpenAI day from the sidelines. Conviction is expensive when the narrative reverses this often.
The round trip ride
A buy and hold investor from January 2024 was up roughly 50% by March, down roughly 45% by April 2025, and back to even only after the OpenAI deal. Two years of holding, most of it spent underwater relative to the peak, for a gain that arrived almost entirely in the final stretch. The account survived. The investor's nerves usually do not.
The too high freeze
By the time AMD proved the 2026 story with earnings, it had already doubled. Every entry point looked like the top. Investors who waited for a comfortable price watched a $214 stock become a $580 stock, then finally bought the July pullback, near the exact moment sellers took over.
Dollar cost averaging deserves honest credit here. It kept buying through April 2025 when nobody wanted to, and it never chased a headline. But DCA has a structural cost: it buys blind, it never sells, and it never realizes a dollar of profit. In a stock this violent, DCA rides every drawdown at full weight and banks nothing at the top.
A rules-based algorithm restores what AMD takes away: a plan that does not renegotiate itself every time the story changes.
Market Wave Algorithm for Equities
The algorithm
The Market Wave Algorithm for Equities does three things, and it does them without asking how you feel about the news.
Rides confirmed moves
It enters when price action confirms a move, not when a headline promises one, and it keeps adding within its rules while the move holds.
Exits on rules, not nerves
It takes profit into strength at levels set in advance. Gains become cash because the system says so, not because the holder finally flinched.
Steps aside in chop
When the wave flattens, it stops trading. No revenge trades, no averaging into a falling knife.
The same algorithm can also be configured for the question you ask it. Scope runs from 0.5, reading the market hour by hour, to 20, reading it in quarters. This study uses three recipes for three jobs: a 0.5 micro scope on hourly bars to scalp a stock trading at all time highs, a mid scope of 5 on daily bars to work through a two year cool off and recovery, and a 14 macro scope on four hour bars to ride this year's run. Same discipline, three different jobs.
The all time high scalp, hourly chart
Scope 0.5
0.5 micro 20 macro
A 0.5 micro scope on hourly bars to scalp a stock trading at all time highs.
The cool off rehearsal, daily chart
Scope 5
0.5 micro 20 macro
A mid scope of 5 on daily bars to work through a two year cool off and recovery.
The getting paid year, four hour chart
Scope 14
0.5 micro 20 macro
A 14 macro scope on four hour bars to ride this year's run.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
Scenario one: the micro recipe
“The all time high scalp”
May 1, 2026 to August 12, 2026 · Hourly chart · Scope 0.5
Call this one "the all time high scalp." May 1 to August 12, 2026, on the hourly chart with scope at 0.5, the finest setting the algorithm offers.
This window answers the question every investor asks a stock like AMD right now: it has already run, so is it too late? The whole window happened at or near all time highs. AMD ran from roughly $352 at the start of May through a record close of $580.91 on June 30, tumbled 18% through July, beat second quarter estimates in early August, and sold off anyway to today's price near $487. Buy and hold captured +38.3% and every point of the July drawdown with it. Dollar cost averaging, buying monthly across the window, earned +7%, its later buys landing near the top.
The micro recipe never needed the run to continue. It scalped confirmed hourly moves, both directions, in percentage sized steps, and banked each one.
Strategy return
+44.3%
Buy and hold +38.3%, outperformance +6 percentage points
DCA +7%, outperformance +37.3 percentage points
Ending value
$144,309
Realized profit, banked in cash
$43,862
Unrealized profit on open shares
$446
Shares held
295 at $485.67 avg cost
Cash
$591
Open full-size chart
The number to read Read the realized number. Nearly all of this window's gain, $43,862 of it, is banked cash, not paper. That is the answer to "is it too high to buy." The algorithm did not bet that a stock at record levels would keep setting records. It extracted 44.3% from three and a half months of topping, pulling back, and churning, and the July slide that erased 18% of a buy and holder's peak value passed through an account that had already converted most of its gain into dollars.
Backtested results. Hourly chart, valued August 12, 2026. Past performance does not guarantee future results. For educational purposes only.
Scenario two: the round trip
“The cool off rehearsal”
January 1, 2024 to January 1, 2026 · Daily chart · Scope 5
Call this one "the cool off rehearsal." January 1, 2024 to January 1, 2026, on the daily chart with scope at 5.
Every AI investor carries the same quiet fear: what if the sector cools off and the market trends down for the next year or two? You do not have to imagine it. AMD already ran that experiment. This window contains the nine week melt up to the March 2024 high, a year long unwind as the AI trade cooled on this exact stock, the April 2025 tariff low near $76, and the slow recovery. A buy and holder who entered near $139 spent most of two years underwater relative to the peak and finished at +54.5% only because the last three months rescued the first twenty one. Dollar cost averaging earned +49.2%, its blind monthly buys through the 2025 low doing exactly what DCA does well.
The algorithm beat them both through the full cycle, and it banked cash the whole way down and back.
Strategy return
+57.9%
Buy and hold +54.5%, outperformance +3.3 percentage points
DCA +49.2%, outperformance +8.7 percentage points
Ending value
$157,866
Realized profit, banked in cash
$66,516
Shares held
600 at $228.58 avg cost
Cash
$29,370
Open full-size chart
The number to read Read the basis. The discipline of buying weakness inside rules left this account holding 600 shares at a $228.58 average cost. AMD trades near $487 today, which makes that same position worth $292,530 as of this writing. That is what a cool off is worth to a rules-based account: the drawdown everyone fears became the accumulation phase, and the recovery paid it in full. If the next AI cool off looks anything like the last one, this window is the plan for it.
Backtested results. Daily chart, window valued January 1, 2026. Position value today from public market data. Past performance does not guarantee future results. For educational purposes only.
Scenario three: the macro recipe
“The getting paid year”
January 1, 2026 to August 12, 2026 · Four hour chart · Scope 14
Call this one "the getting paid year." January 1 to August 12, 2026, on the four hour chart with scope at 14, configured to ride the big wave and ignore the noise.
This is the 2026 scoreboard, the window everyone will remember: $214 to a $580.91 record close by June 30, the 18% July tumble, the post earnings selloff. Buy and hold earned a remarkable +120.2% and deserves credit for it. Dollar cost averaging, buying monthly across the window, earned +68.9%, a strong year by any normal standard.
The macro recipe did not predict any of it. It rode what the market confirmed, in size, and stepped off pieces on the way up.
Strategy return
+145.3%
Buy and hold +120.2%, outperformance +25.1 percentage points
DCA +68.9%, outperformance +76.3 percentage points
Ending value
$245,267
Realized profit, banked in cash
$141,782
Unrealized profit on open shares
$3,486
Shares held
496 at $480.64 avg cost
Cash
$3,383
Open full-size chart
The number to read Read the realized number again. $141,782 in banked profit is more than the entire starting account. In the best AMD year anyone has seen, the algorithm did not merely keep up with buy and hold. It beat it by 25 points and converted the win into cash while the July selloff was still finding out who was overexposed.
Backtested results. Four hour chart, valued August 12, 2026. Past performance does not guarantee future results. For educational purposes only.
Side by side
AMD results by scenario
The micro recipe
The round trip
The macro recipe
Window
May 1 to Aug 12, 2026
Jan 1, 2024 to Jan 1, 2026
Jan 1 to Aug 12, 2026
Chart and scope
Hourly, 0.5
Daily, 5
Four hour, 14
Strategy return
+44.3%
+57.9%
+145.3%
Buy and hold return
+38.3%
+54.5%
+120.2%
Edge vs buy and hold
+6 pts
+3.3 pts
+25.1 pts
DCA return
+7%
+49.2%
+68.9%
Edge vs DCA
+37.3 pts
+8.7 pts
+76.3 pts
Ending value
$144,309
$157,866
$245,267
Realized profit
$43,862
$66,516
$141,782
Shares held at end
295
600
496
Average cost basis
$485.67
$228.58
$480.64
Cash at end
$591
$29,370
$3,383
All figures from a hypothetical $100,000 starting in cash. 2026 windows valued August 12, 2026. The two year window valued January 1, 2026.
Look at the realized profit row before anything else. In all three windows, in three different market regimes, the algorithm converted movement into banked cash: $43,862 in one quarter, $66,516 across a round trip, $141,782 in the year of the run. Buy and hold produced paper. The algorithm produced dollars, and the dollars did not give anything back in July.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
Why the results differ
The three scenarios are not three strategies. They are one algorithm answering the three questions investors actually ask.
Ask it whether a stock at all time highs can still pay you
Answer: the micro recipe answers: +44.3% scalped from a single topping quarter, banked before the July slide could take it back.
Ask it what happens if the AI trade cools off for a year or two
Answer: the mid recipe answers: +57.9% through the last full cool off, plus a 600 share position at a $228.58 basis handed to the next bull market.
Ask it what this year was worth
Answer: the macro recipe answers: +145.3%, beating a +120.2% buy and hold in the exact conditions where buy and hold is hardest to beat.
The benchmarks could not adapt because they only ever make one decision. Buy and hold decides once and hopes. DCA decides once and repeats. Both were at the mercy of which chapter of the AMD story their one decision happened to land in. The algorithm made hundreds of small decisions, every one of them by rule, and none of them required knowing which story came next.
The edge is not prediction. The edge is never needing to be right about the narrative.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
The real lesson
AMD is the vehicle. The lesson travels.
01
All time highs are not a barrier, they are a market
You do not need a stock to keep setting records to get paid at record levels. The micro recipe scalped +44.3% out of a quarter spent topping and pulling back, and banked it.
02
The cool off you fear is a scenario the rules have already rehearsed
If the AI trade goes cold and the market trends down for a year or two, the round trip shows the plan: keep buying weakness inside rules, keep banking strength, and hand the recovery a large position at a low basis.
03
Banked profit is the only profit that ignores the news
Realized cash, $43,862 in a quarter, $141,782 in a year, does not care what the next headline says. The question is not whether the algorithm works. The question is which job you want it doing in your account.
That last question is the client decision. Bring us the job. The Concierge Team will show you the recipe.
Backtested results. Past performance does not guarantee future results. For educational purposes only.
A path forward
Your capital never leaves your account. The algorithm trades under rules you set, in a brokerage account you own, with zero custody handed to anyone. You keep full control and full transparency. What changes is the decision making: rules, not reactions.
Thirty months of AMD asked investors to be right six times in a row. The algorithm was not right about anything. It rode a two year round trip to +57.9%, worked one violent quarter to +44.3%, and turned the year of the run into +145.3% with $141,782 of it already cash. The buy and holder needed the story to end well. The algorithm got paid while the story was still being argued.
The next stock will change its story too. What is your plan for the day it does?
Book a demo Or talk with the Concierge team
Buy and hold, the getting paid year
+120.2%
The Market Wave Algorithm for Equities, the getting paid year
+145.3%
$141,782 banked as realized profit
More case studies
META
Volatility Harvesting for Meta
Strategy return +217.8% in “The AI race accumulator”
Read the case study PLTR
Volatility Harvesting for Palantir
Strategy return +312% in “The gap buyer, finer wave”
Read the case study MU
MU: a trillion-dollar run, a 41% drop in five weeks, and three recipes that paid in cash
Strategy return +461.7% in “The 2026 scoreboard”
Read the case study
See every case study
Important disclosures
Educational purposes only. Not investment, tax, or legal advice. Consult a qualified professional. Trading stocks, ETFs, cryptocurrencies, and futures involves significant risk, including possible loss of principal. Backtested performance is hypothetical and does not represent actual trading. Algorithm performance varies with settings and market conditions. Past performance does not guarantee future results. Arch Public never takes custody of your assets.
© 2026 Arch Public. All rights reserved. Published August 13, 2026.
Terms and conditions
Privacy policy
Cookies and trackers
Your privacy choices