How to Build a Portfolio: 80% Machine, 20% Hunt

The Scoreboard Nobody Frames

In 2025, 79% of professional US large-cap fund managers failed to beat the S&P 500. Not amateurs - professionals, with research staffs and terminals, paid specifically to do better than an index that requires no skill to own. That figure comes from S&P's SPIVA scorecard, which has been auditing this exact match-up for 25 years. 2025 was the fourth-worst year for the professionals in the study's history.

Hold that number, because every portfolio decision worth making flows from it.

Building a portfolio comes down to two jobs that are usually mashed together: the part of your money that should ride the whole market, and the part - if any - that tries to beat it. The industry calls the split core-satellite. The plain-language version: a machine and a hunt.

We used Barebone to pull the records for both. Ten years of leaving an index completely alone. And the full post-ChatGPT price history of this era's two defining supply-squeeze trades - plus the one nobody puts on a poster.

The receipts: $10,000 became $39,354 without a single trade. NVIDIA returned +1,093% over the same window in which the index did +75%. And Novo Nordisk lost 29% while riding the exact same demand wave that made Eli Lilly a +156% winner.

All three numbers belong in the same portfolio lesson. Here's how they fit.

The 80%: A Machine That Asks Nothing of You

Passive investing means you stop trying to pick winners and buy the entire market in one instrument - an index fund, a single ticker that holds hundreds of companies in market-set proportions. The standard example is SPY, which tracks the S&P 500: roughly five hundred of the largest US-listed companies in one wrapper.

A correction worth making early, because the marketing writes checks the fine print doesn't cash: QQQ, the other fund beginners hear about, is often pitched as "the top 100 tech companies in the world." It isn't. It tracks the Nasdaq-100 - the hundred largest non-financial companies listed on the Nasdaq exchange. Growth-heavy, tech-tilted, but a slice of one exchange, not a clean technology bet. Know what the wrapper actually holds before you treat it as a theme.

What does the boring machine actually produce? Here is the previous full decade, dividends reinvested, no action taken at any point:

Ten thousand dollars invested at the start of 2016 finished 2025 at $39,354 - a compound return of 14.7% per year. No research, no timing, no decisions after day one. Along the way it absorbed a -4.2% year in 2018 and a brutal -18.0% year in 2022, when the position dropped from $25,897 to $21,225 and stayed underwater for over a year.

That's the honest shape of the machine. It compounds impressively on average and looks broken roughly one year in four. The decade above was also a generous one by any standard - the next ten years carry no obligation to repeat it.

The price of admission isn't money. It's sitting still through 2022 without touching anything.

Paying In: What Monthly Buying Actually Buys

The standard advice for feeding the machine is dollar-cost averaging - DCA - investing a fixed amount on a fixed schedule regardless of what the market is doing, so your purchase prices average out across highs and lows.

Here's the part the advice usually skips. Vanguard tested cost averaging against simply investing everything immediately, across nearly five decades of global market data (1976 - 2022). Investing the lump sum at once won 68% of the time. The logic is mechanical: markets rise more years than they fall, so cash waiting on the sidelines usually costs more than the bad timing it protects against.

So why does almost every serious practitioner still endorse DCA? Two reasons.

First, the study assumes you have a lump sum. Most people don't - they have a salary. Money arrives monthly, so investing it monthly isn't a strategy choice; it's just refusing to let cash pile up. DCA is the natural shape of saving.

Second, DCA deletes a decision. "Is now a good time?" is the question that keeps people in cash for years - and the data says the waiting itself is the expensive part. A standing monthly buy converts an unanswerable timing question into a calendar entry.

Cost averaging doesn't maximize returns - the math mildly favors investing sooner. What it maximizes is the odds you actually keep showing up. On a ten-year horizon, that's the variable that pays.

The 20%: Hunting the Gap

If the machine compounds at market rates, the only reason to pick individual stocks is to do meaningfully better - and the cleanest framework for where that happens is a supply-demand gap: demand arrives faster than supply can physically respond, and the companies controlling the scarce thing re-rate.

The defining example began on November 30, 2022, the day ChatGPT launched. Demand for AI compute went vertical almost immediately; the supply side - advanced chips, data centers, the power to run them - takes years to build. NVIDIA sat at the choke point. Its split-adjusted close on launch day was $16.89. On April 17, 2026 it closed at $201.45 - up +1,093%, nearly twelve-fold, against +75% for the index.

The same mechanism ran in pharmaceuticals. Demand for GLP-1 weight-loss drugs outran manufacturing so badly that the FDA kept semaglutide (Ozempic, Wegovy) and tirzepatide (Mounjaro, Zepbound) on its official drug-shortage list for more than two years - tirzepatide wasn't declared resolved until December 19, 2024, semaglutide until February 21, 2025. Eli Lilly, the company best positioned to close that gap, went from $361.14 to $925.44 over the ChatGPT window: +156%.

Name	The gap	Nov 30, 2022	Apr 17, 2026	Change
NVIDIA	AI compute vs chip/power supply	$16.89	$201.45	+1,093%
Eli Lilly	GLP-1 demand vs manufacturing	$361.14	$925.44	+156%
S&P 500	(benchmark - the machine)	4,080.11	7,126.06	+75%
Novo Nordisk	the same GLP-1 gap	$57.11	$40.52	-29%

That fourth row is the most important number in this article.

The Same Wave Drowned the Other Swimmer

Novo Nordisk is not an obscure also-ran. It invented the GLP-1 trade - Ozempic and Wegovy were its drugs, first to market, the very products whose scarcity put semaglutide on the FDA shortage list. If "find extreme demand, find constrained supply" were a complete framework, Novo should have been its best advertisement.

For a while it was. The stock peaked on June 25, 2024 at $137.40. By April 17, 2026 it closed at $40.52 - down 70% from the peak, down 29% from ChatGPT day, after losing roughly half its value in 2025 alone.

Nothing about the demand thesis failed. What failed was everything the simple framework ignores. Lilly's tirzepatide turned out to be the stronger product. Novo's next-generation drug, CagriSema, disappointed in successive trial readouts - capped in February 2026, when a head-to-head study failed to show it was even non-inferior to tirzepatide and the stock fell 16.5% in a day to its lowest level since June 2021. Compounding pharmacies legally flooded the market with copies during the shortage years. Management cut guidance repeatedly through 2025. And the entry price in 2024 already assumed permanent leadership, so every piece of bad news repriced the whole story.

The framework found the right wave twice. It still produced one of the decade's best trades and a three-year loss - because a gap everyone can see is a gap the market has already priced. The variable that decided the outcome wasn't the wave. It was which swimmer you picked, and what you paid.

This is why the base rates from the top of the article matter. Stock picking has a measurable difficulty curve, and it's measured on people who do it full time:

Give the professionals one year and 79% lose to the machine. Give them fifteen and 89.5% do. Retail does worse: Barber and Odean's classic study tracked 60,000-plus households at a discount brokerage and found the average self-directed investor earned 15.3% a year net of costs against the market's 17.1% - while the most hyperactive fifth earned 10.0%, surrendering roughly seven points a year to activity itself.

Active investing isn't forbidden by the data. It's just expensive to do casually. Which is an argument about size.

Why 80/20 and Not 50/50

The 80/20 split isn't mystical. It's loss-containment arithmetic with the upside left intact.

Allocation	If the active sleeve went to zero	If it had caught NVIDIA's run
100% passive	0 impact	+0 points
80/20	-20 points	+219 points
50/50	-50 points	+546 points

A 20% sleeve that fails completely costs you about as much as one bad index year - 2022 was -18% - which the machine itself recovered from within two years. A 20% sleeve that catches a genuine gap trade adds roughly 219 points of whole-portfolio return (0.20 × 1,093%), most of what the machine needed an entire decade to produce.

A 50/50 split doubles the dream scenario, but its failure case is half the portfolio - a hole that requires a clean double just to get back to even, inflicted by your own picks rather than by a market crash. The asymmetry you want is a capped, survivable downside against an uncapped upside. 80/20 buys exactly that.

Should the ratio ever move? Yes - on receipts, not on confidence. If your active sleeve has a multi-year written record beating the index, the data supports giving it more room. If it doesn't, the SPIVA chart above is the prior: the longer the horizon, the more the burden of proof rises, and the default drift should be toward the machine. Feeling smarter after a winning year is not a record. 2025 minted a lot of people who felt smarter.

What This Means

Five working rules fall out of the data:

Build the machine first. The 80% is the only part of a portfolio backed by a 25-year audit trail showing that most professionals can't improve on it. It does the compounding; everything else is attempted alpha.

Let the contributions be boring. Fixed amount, fixed date. The lump-sum study says waiting in cash for a better entry costs more than it saves about two-thirds of the time. The discipline is the edge; the timing rarely is.

Hunt only gaps you can write down. Named demand, named supply constraint, and a reason the constraint survives the next two years. NVIDIA cleared that bar in 2023. If you can't name what's scarce, it isn't a gap trade - it's a mood.

Watch the supply side for the ending. Gaps close. The FDA shortage list clearing in early 2025 was the GLP-1 gap formally ending; competitor capacity and copied products were Novo's repricing in real time. The demand headline is the ad. The supply detail is the trade.

Size for being wrong, not for being right. The 20% cap is what makes the inevitable Novo in your future a lesson instead of a setback.

The market pays for patience and it pays for being right early - and it pays patience far more often. An 80/20 portfolio is just that sentence, written as an allocation.

---

Data: Barebone | Sources: SPIVA U.S. Scorecards (S&P Dow Jones Indices, year-end 2024 and 2025), Vanguard, "Cost averaging: Invest now or temporarily hold your cash?" (2023), Barber & Odean, "Trading Is Hazardous to Your Wealth", FDA drug-shortage resolution orders (December 2024, February 2025) | Data as of April 20, 2026