The Portfolio Your Search Bar Built — What If You Picked Stocks by Search Volume Alone?

2021’s most-looked-up stocks: if you’d bought them then, where would you be now?

We usually pick stocks by looking at earnings and charts. But what if you skipped all of that and built a portfolio only from “the stocks people look up the most”?

It sounds like a reasonable experiment. Gather today’s popular stocks, run their past returns, and the results can be stunning. But there is a very common trap here. If you pick today’s popular stocks and go back in time, you have effectively chosen stocks you already knew had gone up.

So this time we did the opposite. Suppose that at the end of December 2021 you had bought the stocks people looked up most that year. And suppose you then held that account, untouched, until September 30, 2026. Did the portfolio your search bar built beat the market?

Starting with the result: the cumulative return was +74.7%. Over the same period the S&P 500 returned +71.1%. On the surface, not bad at all.

Open up the portfolio, though, and the story changes completely. NVIDIA alone produced +85.0 percentage points, while the other seven stocks together came to −10.3 points.

We took the numbers apart to see what following “the stocks drawing the most attention” really means.

When people search for a stock a lot, it shows at least one thing: people are paying attention to it. The problem is that attention and investment performance are not the same thing.

Da, Engelberg and Gao (2011) showed that a surge in search volume can serve as a measure of stock-market attention, and that it can be linked to short-term price moves, especially in stocks that retail investors watch closely. Barber and Odean (2008) likewise showed that individual investors tend to notice and trade the attention-grabbing stocks first among thousands of choices.

The message the two studies share is simple.

A stock that many people look at is not, for that reason, a good stock.

If anything, stocks that have already made big price moves tend to draw more attention. So what happens if you turn that attention into an actual portfolio? The most important question, though, is when that information could have been known.

Taking 2026’s popular stocks and computing their returns from 2021 is not a fair experiment. It amounts to carrying stocks that are popular today, the survivors, back into the past and investing in them. This is called look-ahead bias.

So this time we used only a ranking that was actually public when 2021 ended: “the top 10 trending tickers of 2021,” published by Yahoo Finance on December 29, 2021.

It is based not on Google search volume but on how many times people viewed each ticker’s page on Yahoo Finance. The No. 1 ticker, AMC, was viewed about 194.9 million times that year.

Of the ten, we left out the two that are not stocks, the Dow Jones Industrial Average and bitcoin, and put the remaining eight individual stocks in the basket.

The rules were kept very simple. At the December 31, 2021 close, $10,000 goes into the eight stocks in equal $1,250 slices. After that, nothing is touched. Dividends are reinvested; taxes, fees and currency effects are left out.

In other words, the experiment asks one question.

“If you had bought trusting only the attention ranking you could have known in 2021, what would your account look like now?”

Result first: the $10,000 became $17,465, a cumulative return of +74.7%. Over the same period the S&P 500 returned +71.1% and the Nasdaq-100 +91.3%.

On the numbers alone, it looks pretty good. “So just picking the most-searched stocks could have earned about what the market did?” But the road this account traveled was completely different.

In 2022 alone the search basket lost 51.6%, a far steeper fall than the S&P 500’s −18.2%. By late December 2022 the $10,000 had shrunk to $4,689.

The principal came back in May 2024, after more than two years and four months under water. The S&P 500, by contrast, had already recovered its principal in December 2023.

Volatility was 31.7%, about twice the market’s, and the Sharpe ratio of 0.40 was lower than the S&P 500’s 0.55.

So it is hard to say only that this basket “earned about what the market did.” More precisely:

To end up with a market-like return, it had to endure far greater volatility and drawdowns.

Now let’s open up the eight $1,250 positions one by one.

This is where the heart of the experiment shows. The search basket returned +74.7% overall, yet NVIDIA alone produced +85.0 points. Add up the contributions of the other seven and you get −10.3 points.

Had you left NVIDIA out and bought only the other seven the same way, the result would have been −11.8%.

So this portfolio’s +74.7% is not a story of “we picked high-attention stocks and they all rose nicely.” It is closer to “one ticket out of eight hit the jackpot.”

Four of the eight lost money. AMC turned $1,250 into $22, and NIO shrank to $135. Tesla essentially went nowhere for four years and nine months.

The median return of the eight was −1.5%. A mean of +74.7%, a median of −1.5%. That gap is the essence of this experiment. Judged by the average alone it looks like a successful portfolio, but most of its holdings did not perform that way.

That raises a question: “Still, NVIDIA was in there, so isn’t that a success?” True. Which is why the more important question is “could you have recognized NVIDIA on December 31, 2021?”

At the time NVIDIA was No. 8 in the ranking. The event that came to symbolize the AI boom we know today had not happened yet: ChatGPT was released about 11 months later, in November 2022.

In other words, nothing in this experiment would have let you judge, from the end-2021 ranking alone, that “of these eight, NVIDIA will rise 680%.”

The excess return NVIDIA produced is better read not as evidence of predictive power in the attention ranking, but as the result of one stock, a huge success in hindsight, happening to be in the mix.

Rebalancing is even more interesting. Every stock started at 12.5%, but as NVIDIA soared, its weight in the untouched account grew to 55.8%.

Had you reset every weight to 12.5% each month instead, you would have kept spreading NVIDIA’s gains into the other stocks. In that case the cumulative return would have been just +12.4%.

Same eight stocks, very different outcomes, depending on whether you “leave the winners alone” or “reset to the original weights.”

So this experiment is not only about stock selection. A portfolio’s performance depends heavily not just on which stocks you picked, but on how their weights changed over time.

Separately from the actual buy-and-hold result, we also checked the basket’s character with DIVA’s Markowitz analysis. For that purpose the model assumes a portfolio kept at 12.5% per stock and rebalanced monthly, so the expected return it reports is not the same number as the +74.7% buy-and-hold result above.

Volatility was about 2.3 times the market’s over the same period. The individual stocks averaged 56.8% volatility, but mixing different industries and companies brought the portfolio down to 35.6%. So the diversification effect itself was clearly there.

But because the starting point, the stocks’ own high volatility, was so high, overall risk stayed substantial even after mixing. On DIVA’s efficient frontier, too, the basket sat well away from the more efficient combinations the same eight stocks could have formed.

DIVA classifies this portfolio’s archetype as “The Mapless Explorer.” It doesn’t literally mean taking risk without direction; it means a basket that took on high volatility but was not efficiently rewarded for that risk.

Finally, we used the Fama–French five-factor model to check what stock style this portfolio had.

Market beta was +1.15: a portfolio that moved a little more than the market. Size beta was +1.13 and value beta −1.02. In factor terms, a strong small-cap growth tilt showed up.

What’s interesting is that this came out even with mega-caps like Apple, Amazon and NVIDIA in the basket. The moves of stocks like AMC, GameStop and NIO pulled the whole portfolio’s style toward small-cap growth.

So was there any special excess return of its own, any alpha? Annual alpha was +3.3%. But the p-value was 0.78, statistically indistinguishable from zero. R² was 58.6%.

Looking at what this basket earned, it is hard to argue there was any special “search alpha.” Market exposure produced the largest share of the return, while the small-cap and growth tilts weighed on results over this period.

Picking high-attention stocks did not, in itself, produce any special excess return.

Here is the experiment in one sentence.

A search ranking is less a source of future investment ideas than a list of stocks people have already been watching.

The most-viewed stocks of 2021 included some that had already soared, like AMC and GameStop, and some that had already crashed, like Alibaba and NIO. That people looked them up means something was going on with those stocks, but it does not mean that something will keep pushing them up.

This basket’s numbers show it well. The average return was +74.7%. But the median was −1.5%. And what made that average was, in effect, NVIDIA alone.

The most important lesson for investors here is not simply “don’t trust search volume.”

First, look at what is hiding behind the average.

A single portfolio return can look like success. Spread out the individual stocks, though, and half of them lost money, with one stock’s jackpot masking the overall picture.

Second, separate the outcome from what was known at the time.

At the end of 2021 there was no way to know NVIDIA would rise like this. The moment you go back in time already knowing today’s winners, the experiment stops resembling a real investment decision.

So when testing an investment idea, always ask first: “What could have been known at the time?”

Third, the path to a return matters as much as the return.

The search basket finished with a return similar to the S&P 500’s, but its maximum drawdown was −54.7%. Even when the results look the same, what the investor has to endure along the way can be entirely different.

Search rankings keep changing. In 2021, AMC, GameStop and NIO grabbed people’s attention. Since then, other stocks have taken their place.

The problem is that when you look at past charts of today’s popular stocks, it is all too easy to make up a story: “If I had bought here, I’d have made this much.” But one question is missing from that calculation: “Back then, could I really have known about this stock?”

So if a stock you’re watching becomes an investment idea, one approach is to record where things stand now. Note which stocks you’re looking at, what the price is, and why you got interested, and a few years later you can compare that with what actually happened.

In DIVA Quantizer you can enter the stocks you’re interested in now and check their volatility, factor exposure, diversification effect and more. Rather than imagining “if only I had bought then” a few years from now, recording today’s portfolio today is a far more honest experiment.

Had you split $10,000 equally across the eight most-viewed stocks of 2021, you would be up +74.7% four years and nine months later, almost exactly the market. But most of that return came from NVIDIA alone. Without it, −11.8%. Four of the eight stocks lost money.

What this shows is not that search volume is useless. People’s attention is clearly important market information. It’s just that a lot of attention and a good investment are two separate questions.

The search bar tells you what people are looking at. It doesn’t tell you why they’re looking, how much the stock has already risen, or how much further it will go.

So if you use the search bar as an investing tool, don’t stop at “what’s popular?”
Go on to “what prices and what portfolio structure is that attention creating?”

Reading last year’s report card and predicting the future are two different things.

Data · attention ranking: Yahoo Finance, “AMC Entertainment is Yahoo Finance’s 2021 top trending ticker” (Dec 29, 2021) · prices: yfinance adjusted close, Dec 31, 2021 – Sep 30, 2026 · AMC adjusted for the APE units distributed on Aug 22, 2022, treated as reinvested in AMC at that day’s APE close of $6.00 · factors: Ken French Data Library, North America FF5, 56 months through Aug 2026 · risk-free rate: US 13-week T-bill, 4.00% annual. References · Da, Engelberg & Gao (2011), In Search of Attention, Journal of Finance 66(5) · Barber & Odean (2008), All That Glitters, Review of Financial Studies. All risk and factor metrics are DIVA Quantizer model outputs.

This article was written with the help of AI.