StrategyInvesting & PortfolioGetting Started17 min readPublished November 18, 2026

Can a Random Stock Portfolio Beat an Index Fund?

One random stock a month is an equal-weight bet plus noise. On the noise alone, our model had 13% of such portfolios trail the index fund by 10%+ in 10 years.

Here is an alternative to an index fund that DIY investors sometimes use. Take the current list of companies in the index you would otherwise buy. Each time you have money to invest, let a random number generator pick one of them, and buy it. Strike anything awkward to own, such as partnerships that send a Schedule K-1, before you draw. When a windfall arrives, split it across several draws so no single company gets too much. Never sell.

The case for it has three parts. Ten random stocks in equal weights are said to smooth returns almost as well as a 500-stock cap-weighted fund over a ten-year horizon. Picking at random is said to avoid the problems of market-cap weighting. And buying stocks directly is said to avoid fund costs, commissions and bid-ask spreads. Each claim gets a different answer. The strategy holds up better than the first claim suggests, because the rule never stops buying.

The short version

Ten random stocks are almost as smooth as the index, and over ten years they can still end up far from it. In a model calibrated to current S&P 500 stocks, ten stocks bought once and held a decade ended anywhere from 48% behind to 71% ahead of the index fund in the middle 90% of runs.1 Buying one random stock every month is a different and much tamer portfolio: after ten years it holds about 107 companies, and in the same model 90% of runs finished between 14% behind and 16% ahead of the fund on the same contributions, before any equal-weight tilt, which the model leaves out.

What the rule cannot do is raise expected return beyond what equal weighting offers. Equal-dollar random picks carry the same tilt toward smaller and cheaper companies as an equal-weight index, and that tilt has led and lagged cap weighting in different decades, sometimes by as much as the noise. Against a total-market fund the fee it saves is about 0.03% a year. I would buy the fund for the core of a portfolio. The random rule is a defensible way to own individual stocks with money whose result you can let drift from the market’s, and it has a real use for taxable investors who will harvest losses.

What the rule gets right

It removes the stock-picking decision. The investor never judges which company is undervalued, and a random draw cannot be swayed by a headline, a favorite product, or a story about a CEO. That makes this rule safer than the more common habit of buying a handful of companies you know well, the subject of Five Stocks Are Not Safer Than an Index Fund.

The first few stocks remove most of the volatility. When I drew random equal-weight portfolios from current S&P 500 members for Is VT the Most Diversified ETF?, the first ten stocks removed 87% of the volatility that diversification can remove. On that measure the ten-stock claim is close to right.

Trading is cheap. Fidelity, Schwab and Vanguard charge no commission on online trades of US stocks and most ETFs,2 so buying one company a month costs no more in commissions than buying a fund.

The portfolio keeps growing. A rule that buys one new name a month is concentrated only at the start, when it holds little money. By the time the account is large, it owns dozens of companies. The ten-stock claim describes a fixed portfolio, which the rule resembles only in its first year.

Volatility and tracking are different risks

The ten-stock rule of thumb goes back to Evans and Archer’s 1968 study of S&P stocks, which found that the dispersion of returns fell quickly as securities were added and then flattened.3 The math behind it is simple. Give every stock a market beta of one and its own company-specific volatility σe\sigma_e, independent of the others. An equal-weight portfolio of NN stocks then has volatility

σp=σm2+σe2N\sigma_p = \sqrt{\sigma_m^2 + \frac{\sigma_e^2}{N}}

where σm\sigma_m is the market’s volatility. Over November 2021 to October 2026, the root-mean-square company-specific volatility of current S&P 500 members was 30.9% a year, and SPY’s volatility was 15.5%.1 Plug in ten stocks and the portfolio’s volatility comes out at 18.3%, about 18% higher than the index’s. That is the smoothing the strategy promises, and it is real.

The same company-specific risk also makes the portfolio drift away from the index, and that drift is what an investor who expected index-like results will notice. Its annual size, the tracking error, is

TE=σeN≈30.9%10≈9.8% a year\text{TE} = \frac{\sigma_e}{\sqrt{N}} \approx \frac{30.9\%}{\sqrt{10}} \approx 9.8\% \text{ a year}

and over TT years it grows at least as fast as T\sqrt{T}, and faster when the portfolio is never rebalanced and its winners take over. A longer horizon makes volatility matter less for average annual returns, and it makes the gap between your ending wealth and the index’s wider. Ten years is a reason for more caution about ten stocks.

To see what that looks like, I simulated random portfolios with exactly those inputs: each stock gets the index’s expected return and beta, so any difference from the index is pure company-specific noise.1

Random stocks, bought once and held 10 yearsVolatility5th pctMedian95th pctTrail by >10%
1 stock34.6%0.12×0.63×3.21×64%
5 stocks20.7%0.39×0.87×2.02×53%
10 stocks18.3%0.52×0.92×1.71×47%
30 stocks16.5%0.68×0.97×1.41×37%
100 stocks15.8%0.81×0.99×1.22×22%
Index fund15.5%1.00×1.00×1.00×0%

Author’s simulation, 10,000 paths per row. Ending wealth is a multiple of the index fund’s ending wealth on the same market path. Every stock has beta 1 and the index’s expected return; company-specific volatility 30.9%, measured on current S&P 500 members, and index volatility 15.5%, measured on SPY, November 2021 to October 2026. Volatility column is the closed-form formula above. Equal dollars at the start, no rebalancing, no costs or taxes.

Two features of the table matter more than the volatility column. First, the median sits below 1.00 at every size even though every portfolio has the same expected ending wealth as the index. A few companies compound into large winners, they pull the average up, and a small portfolio usually misses them. Second, ten stocks trailed the index by more than 10% in close to half the runs and by more than 25% in more than a quarter. Holding the same ten for 30 years made it worse: in the same model the median ten-stock portfolio ended at 0.69 times the index and 54% of runs trailed it by more than a quarter.

The published research says the same thing in other ways. Domian, Louton and Racine studied portfolios drawn from 1,000 large US stocks over a 20-year holding period and found that the chance of ending below a target kept falling even above 100 stocks.4 Alexeev and Tapon found that across five developed markets, reaching a diversification standard most of the time, rather than on average, took more than 73 stocks.5 In Bessembinder’s study of US stocks from 1926 to 2016, random portfolios of five stocks trailed the value-weighted market over ten-year horizons in 59.2% of simulations, and portfolios of 100 stocks did so in 52.5%; his portfolios were redrawn every month and weighted by company size within the portfolio.6 His 2026 study of “do-nothing” portfolios drawn from S&P 500 members since 1971, which keep every stock even after it leaves the index, reports that narrow random portfolios “generate average returns similar to the index, but underperform the majority of the time”. Weighted by market value and left untouched for ten years, his random portfolios beat the index in 41.5% of draws with one stock, 45.0% with five and 49.3% with 100.7

Company-specific volatility also moves around. Campbell, Lettau, Malkiel and Xu found that it spiked in 1999 and 2000, declined afterward, and rose again in 2008 to 2009 and in 2020.8 A five-year calibration is a snapshot, and current index members are survivors, which tends to understate it.

One stock a month: how the portfolio grows

The rule as practiced never stops buying. With MM companies in the universe and kk random draws, allowing repeats, the expected number of different companies owned is

E[Dk]=M(1−(1−1M)k)E[D_k] = M\left(1 - \left(1 - \tfrac{1}{M}\right)^{k}\right)

For a 500-company universe that is about 12 companies after a year of monthly buys, 57 after five years, 107 after ten and 257 after thirty. Repeats slow the count down: by year ten, about 13 of the 120 draws landed on a company already owned.

Company count overstates diversification when weights are unequal, so the better measure is the effective number of holdings,

Neff=1∑iwi2N_{\text{eff}} = \frac{1}{\sum_i w_i^2}

which equals the count when every weight is equal and shrinks as a few positions grow. Because the rule never sells, early purchases that do well keep compounding. In the model, the median effective number reached 57 at year ten, peaked at about 59 around year thirteen, and then fell to 41 by year thirty even though the portfolio by then owned 257 companies.1 For comparison, the S&P 500’s own effective number was about 46 in our idiosyncratic risk measurement. The difference is that the index’s concentration is the market’s, so it can never make you trail the market.

One random stock a month for5th pctMedian95th pctTrail by >10%Fund fee avoided
10 years0.86×0.99×1.16×13%0.17%
20 years0.81×0.98×1.24×24%0.37%
30 years0.75×0.96×1.36×33%0.61%

Author’s simulation, same model as the table above: 500 companies, $1 a month into one random company or into the index fund, ending wealth as a multiple of the fund’s. 10,000 paths at 10 years, 5,000 at 20 and 30. “Fund fee avoided” is the share of the fund’s ending wealth that a 0.03% expense ratio costs over the same contributions.

Ten years of monthly buying ended closer to the index than 100 stocks bought once and held for the decade, and far closer than ten stocks, partly because the average dollar was invested for only about half the decade. The fee avoided, 0.17% of ending wealth, is less than a hundredth of the 30-point gap between the 5th and 95th percentiles. The rule’s mean ending wealth matches the fund’s in this noise-only model, so the spread is risk taken for no expected reward: what Compensated vs. Uncompensated Risk calls an uncompensated risk.

Three checks on the model. Giving each stock its own measured beta and volatility, with expected returns set by the CAPM, widened the ten-year range to 0.84 to 1.17 times the fund, with 20% of runs trailing by more than 10%. Swapping normal shocks for fat-tailed ones changed little. And when I drew random ten-stock portfolios from the actual returns of current members over the past five years, the spread of outcomes was slightly wider than the model’s (a standard deviation of log relative wealth of 0.26 against 0.24), and for 100 stocks the two matched (0.077 against 0.076).1

What the model leaves out on purpose is the equal-weight tilt covered in the next sections, and that tilt can be as large as the noise. As a gauge, $1 a month into RSP, the equal-weight S&P 500 fund, instead of SPY over the ten years to October 2026 ended at 0.84 times the SPY account. Across all 162 overlapping ten-year windows since 2003 the RSP account ended between 0.78 and 1.11 times the SPY account, and below 0.90 in 32% of them.1 Those windows all end in 2013 or later, a stretch that favored cap weighting. The random rule carries less of this tilt than RSP, because it never rebalances its older holdings, but the tilt adds to the noise in the tables above.

Simulate your own version

Set how long you buy, how often, how many companies each purchase is split across, and how volatile the companies are. The defaults match the ten-year row above to within the sampling noise of 1,000 paths.

Random selection is an equal-weight bet

Everything above deliberately gave every stock the index’s expected return, to isolate the noise. Real random picks also carry a systematic tilt. If each of MM companies is equally likely to be drawn and each draw gets the same dollars, every company’s expected share of the dollars you put in is

E[wi]=1ME[w_i] = \frac{1}{M}

which is what an equal-weight index holds. Each purchase therefore earns, in expectation, the equal-weighted average return of the universe from its purchase date on. Because nothing is rebalanced, older purchases then drift toward weights set by their own performance. The arithmetic Sharpe laid out for active management applies here: before costs, the cap-weighted market return is the average return of every dollar invested, so a portfolio weighted differently can beat it only if other investors trail it by the same amount.9 Equal weighting can still win for long stretches when smaller and cheaper companies outperform, since it holds them at the same weight as the giants.

The strongest evidence for the claim that random picking avoids cap weighting’s problems comes from two “monkey” studies. Arnott, Hsu, Kalesnik and Tindall built 100 “monkey” portfolios, each 30 stocks drawn at random from the largest 1,000 US companies, equally weighted and redrawn every year from 1964 to 2012. They averaged 11.26% a year against 9.66% for the cap-weighted portfolio, before costs, and 96 of the 100 matched or beat it.10 A Cass Business School team ran 10 million random-weight versions of the largest 1,000 US stocks, rebalanced yearly from 1969 to 2011, and found that nearly every one beat the cap-weighted index.11

Both studies also explain the result, and the explanation limits what it means for the rule. Arnott and coauthors found that the monkeys loaded on the size and value factors (0.37 and 0.13), and that a four-factor model left them an alpha of −0.29% a year (t = −0.31). Exposure to small and cheap stocks accounted for essentially all of the edge.10 The Cass monkeys beat cap weighting in only 60% of rolling three-year windows, and cap weighting won every window in 1972 to 1975 and 1996 to 2001.11 Each Cass portfolio also held hundreds of names, so it carried the tilt with almost none of the noise in the tables above.

The closest published analogue to the rule itself is Bessembinder’s 2026 study of S&P 500 members from 1971 to 2025. Equal weighting beat value weighting over the period, 12.59% against 11.30% a year with annual rebalancing. Random equal-weighted portfolios left untouched for ten years beat the index in 54% to 56% of draws with 25 to 100 stocks, and in 49.0% with five, before trading costs. The edge shrank the longer positions were left alone: a dollar grew to $678 against $361 with annual rebalancing, but to $387 against $342 when portfolios were reset only every ten years, partly because smaller companies delisted more often and some grew into large ones.7

The live equal-weight fund tells a more mixed story. RSP, the equal-weight S&P 500 ETF, resets every company to the same weight each quarter and charges 0.20% a year.12 From June 2003 to October 2026 it returned 10.9% a year against 11.2% for SPY, net of each fund’s fees. It returned 8.4% a year against 6.2% from mid-2003 through 2012 and trailed slightly from 2013 through 2022. From 2023 through 2025 it gained 42.6% while SPY gained 85.5%, and from January to October 2026 it led again, 15.3% to 11.7%. Over all 162 rolling ten-year windows, which overlap heavily and all end between 2013 and 2026, RSP finished ahead in 46%.1 Over its full history RSP loaded +0.207 on value, +0.106 on size and −0.104 on momentum, as measured in Is the S&P 500 a Momentum Trade?

A random rule that never sells differs from RSP in one useful way. RSP trims its winners every quarter; the random rule lets old holdings drift with their own performance, the way cap weighting does, and puts only new money into equal-dollar picks. Its tilt toward smaller companies lives mainly in its recent purchases, and it never pays for rebalancing trades.

Why there is no backtest of today’s index members here

The tempting test is to take today’s S&P 500 list and see how random picks from it did. Over November 2021 to October 2026, an equal-weight portfolio of the 492 current constituent stocks with five years of history beat SPY by 7.5% in total. RSP, which held the actual members on each date, trailed SPY by 14.5% over the same months. Rebalancing today’s members quarterly, as RSP does, still beat SPY by 6.8%, and RSP’s fee explains about a point, so most of the gap comes from the list itself.1 Today’s list includes companies that joined after rising and leaves out the ones that fell out, so a backtest built from it flatters any strategy that buys smaller members. That is why the simulations here use a model, and why the historical evidence comes from studies built on point-in-time databases.

Compared with a factor fund

If the goal is to own fewer mega-caps, an equal-weight fund does that directly with all 500 companies and no company-specific noise, at the cost of a higher fee and quarterly turnover; RSP reports no capital gains distributions since its 2003 launch.12 If the goal is the size or value premium, a systematic factor fund targets it on purpose. It screens for the characteristics behind the premium, such as low price relative to book value and high profitability, holds hundreds of names, and controls how far it strays from the market. Equal weighting picks up size and value only as a side effect of holding every small company at the same weight, whatever its price or profitability. The ABCs of Systematic Factor Investing covers how those funds are built, and Is Factor Investing Dead? covers what the premiums have done since they were published.

Market-cap weighting has its own case. It is the only portfolio every investor can hold at once, it trades only when the index changes, and it lets winners grow without selling them. Its weakness is that it concentrates in whatever has risen most; on November 17, 2026, the ten largest companies made up 38.6% of SPY.13

Costs: commissions, spreads, and the fee you save

Cost sat at the center of Statman’s 1987 answer. He weighed the risk reduction from each added stock against the extra cost of an index fund and concluded that a diversified portfolio needed at least 30 stocks for an investor who borrows and 40 for one who lends.14 When he returned to the question in 2004, his abstract read: “Today’s optimal level of diversification, measured by the rules of mean–variance portfolio theory, exceeds 300 stocks.”15 Index fund fees have fallen further since. VTI, now the Vanguard Morningstar Total Stock Market ETF, charges 0.03% a year, which is $30 on $100,000.16 If the equal-weight tilt is what you want, the comparable fund is RSP at 0.20%, and there the rule saves more: in the model’s ten-year monthly run, a 0.20% fee costs about 1.1% of ending wealth, and the rule also skips RSP’s quarterly rebalancing trades.1

Commissions are zero for both choices at the large brokers.2 Spreads apply to both as well. Every exchange-traded security, stock or ETF, trades with a gap between the bid and the ask, so buying one stock instead of one fund does not remove it.

One practical cost is easy to miss. At the end of October 2026, 13 S&P 500 companies traded above $1,000 a share and 43 above $500, with the highest near $6,000.1 Monthly purchases of a few hundred dollars need fractional shares. Fidelity and Schwab sell fractions of most US-listed stocks from $1; Vanguard’s dollar-based trading covers its own ETFs and mutual funds but not individual stocks.2 A rule that skips companies it cannot afford is no longer random.

Taxes: what direct ownership adds and what it costs

Owning the companies directly gives a taxable investor control an ETF holder lacks. You can sell individual positions at a loss to offset gains elsewhere while keeping the rest, give your most appreciated lots to charity (see donating stock or using a donor-advised fund), and choose which lots to sell. Chaudhuri, Burnham and Lo estimated that harvesting losses among the 500 largest US stocks from 1926 to 2018 was worth 1.08% a year before trading costs, or 0.82% with the wash-sale rule applied, at assumed tax rates of 35% on short-term and 15% on long-term gains.17 Those benefits shrink as positions rise above their cost, and they defer tax more than they eliminate it; Do You Actually Need Direct Indexing? and Harvested Losses Are Not Savings work through both. A random rule that keeps adding new lots does keep creating fresh positions to harvest.

Direct ownership also creates tax events a fund absorbs:

  • Cash takeovers force a sale. When a company you own is bought for cash, you have transferred your shares for money, which the IRS treats as a sale, on the buyer’s schedule.18 An index fund meets the same deal inside the fund. The SEC notes that “ETFs can be tax efficient because many ETFs buy and sell portfolio securities in in-kind exchanges”,19 and VTI has paid no capital gains distribution in the past ten years.16
  • Random buys can collide with harvested losses. A loss is disallowed if you buy the same stock within 30 days before or after the sale, including a purchase in your IRA or by your spouse.1820 If the generator draws a company you sold at a loss in the past 30 days, skip it. Before harvesting a position, check that neither a draw nor a reinvested dividend bought more of it in the previous 30 days; a reinvested dividend is a purchase too.
  • Every purchase is a tax lot. Twelve lots a year, plus spin-offs and mergers that split or convert positions, make records longer. How Often Should You Invest? covers the lot problem from frequent buying.

The K-1 problem barely arises. The major US index families exclude limited partnerships, including master limited partnerships, from eligibility, so a current constituent list of the S&P 500, the Russell indexes or the Morningstar US indexes contains no partnerships to strike.212223 The more common trap is share classes: the S&P 500 held 503 securities for 500 companies on November 17, 2026, because Alphabet, Fox and News Corp each appear twice.13 Drawing from the ticker list gives those three companies double odds.

How to run the rule without fooling yourself

If you use the rule, write it down first and follow it mechanically.

  1. Draw from a list of companies, with each company once, rather than a list of tickers.
  2. Decide every exclusion before the draw. Redrawing after you see a name you dislike turns the rule back into stock picking.
  3. Use a generator you cannot steer and keep a log of each draw, so you can check later that you followed it.
  4. Decide in advance what happens when a holding is acquired (put the cash into new draws) or leaves the index (keep it, as Bessembinder’s do-nothing portfolios do).
  5. Compare the account with what the same contributions would have earned in an index fund, so you can see the tracking gap rather than guess at it.

Who should use it, and who should not

Your situationWhat fits better
Building a first portfolioA broad index fund. One holding, no records, and no gap from the market to explain to yourself during a bad stretch.
Large taxable account, will harvest losses and give appreciated shares to charityThe random rule can work, with the tax rules above. Direct indexing does the same job with tracking control, for a fee.
Worried about mega-cap concentrationAn equal-weight fund expresses that view with all 500 companies and none of the noise.
Want the size or value premiumA diversified factor fund that targets it deliberately.
Already hold a lot of employer stockAn index fund. More single-company positions add to the risk you already carry.
Enjoy owning companies and can live with a gap from the marketThe random rule as a satellite next to an index core, judged against the fund after ten years. Its tax benefits need a taxable account.

What I recommend

Hold the core of your stock allocation in a broad, low-cost index fund. It earns the market’s return with almost no tracking gap for about 0.03% a year, and a global fund adds the rest of the world.

If you want to own individual companies, the random rule is one of the more sensible ways to do it, because it takes stock picking out and keeps buying until the portfolio is broad. Its tax advantages exist only in a taxable account and only if you harvest losses; in an IRA it is the same tracking risk with no tax benefit. In our model it landed within about 15% of the index, on either side, in nine of ten ten-year runs before any equal-weight tilt, and drifted further over longer horizons as winners took over. Do not adopt it to save fees or trading costs, or to beat the index: the savings against a 0.03% fund are tiny, and the expected return is whatever the equal-weight tilt it carries delivers.

Frequently asked questions

Can a monkey throwing darts beat the market?

Random-weight portfolios of large US stocks beat the cap-weighted market from the 1960s to the early 2010s in two large studies, before costs, because they held smaller and cheaper companies at higher weights than the index did. Among S&P 500 members from 1971 to 2025, equal-weighted random portfolios of 25 or more stocks left alone for ten years beat the index in most draws. That edge came from factor exposure and failed in several multi-year stretches. The equal-weight S&P 500 fund trailed the cap-weighted one by a wide margin from 2023 through 2025, then led in the first ten months of 2026. A dart-thrower with only a few darts also takes on company-specific risk the studies largely diversified away.

How many random stocks do I need to match an index fund?

No number closes the gap completely. In the model above, 100 random stocks bought once and held ten years still ended more than 10% away from the index, in one direction or the other, in about 42% of runs. For the narrower question of how many stocks it takes to reduce volatility, see Is VT the Most Diversified ETF?

Is equal weight better than market-cap weight?

In studies covering the 1960s through 2025, equal weight earned more than cap weight over the full periods, with more volatility, and the extra return lined up with exposure to smaller and cheaper companies. Since RSP launched in 2003 it has earned slightly less than SPY, and over shorter periods either can lead by a wide margin. Treat equal weight as a factor tilt you choose on purpose.

Is buying individual stocks cheaper than buying an ETF?

Barely. Commissions are zero for both at the large brokers, both have spreads, and the fee you avoid on a 0.03% fund is $30 a year per $100,000. Ten years of monthly buying avoids about 0.17% of ending wealth in fees, in the model above, while 90% of runs ended between 14% behind and 16% ahead of the index, before any equal-weight tilt.

How is this different from direct indexing?

Direct indexing services also hold individual stocks, but they choose the holdings and weights to track an index closely and harvest losses systematically. They control the tracking error that the random rule ignores, and they charge for it.

I already own a pile of random stocks. Should I sell them?

In a retirement account you can switch to a fund at no tax cost. In a taxable account, selling winners triggers gains, so a common approach is to point new contributions at an index fund, harvest positions that are below cost, and give appreciated shares to charity when you give.

Key takeaways

  • Ten random stocks are nearly as smooth as the index (18.3% against 15.5% volatility at current S&P 500 inputs) and can still end a decade far from it: in our model, close to half of ten-stock portfolios trailed the index fund by more than 10% after ten years.
  • Buying one random stock a month is much closer to the index. In the same model it owned about 107 companies after ten years and finished within 14% behind to 16% ahead of the fund in 90% of runs, before any equal-weight tilt.
  • Without selling, winners take over: the model’s effective number of holdings peaked around year thirteen and fell as the portfolio aged.
  • Random equal-dollar picks carry an equal-weight tilt toward smaller and cheaper companies, which has led and lagged cap weighting by turns, at times by as much as the noise.
  • The cost case is weak against a 0.03% fund. The tax case is real for investors who harvest losses and give appreciated shares, after allowing for forced sales in takeovers and wash-sale collisions.

Related guides

Sources

  1. Author’s calculations. Calibration, simulations, the RSP and SPY comparison and the survivorship check, with code and output, are in summitward-research/random-stocks-vs-index-fund. See the method note below.
  2. Fidelity, Commissions and margin rates and Fractional shares; Schwab, Pricing and Fractional shares; Vanguard, Investment fees and What is dollar-based investing. All read November 18, 2026. Fidelity charges a fee on a limited list of ETFs.
  3. Evans, J. L., and S. H. Archer (1968). Diversification and the Reduction of Dispersion: An Empirical Analysis. Journal of Finance 23(5), 761–767.
  4. Domian, D. L., D. A. Louton, and M. D. Racine (2007). Diversification in Portfolios of Individual Stocks: 100 Stocks Are Not Enough. The Financial Review 42(4), 557–570.
  5. Alexeev, V., and F. Tapon (2014). The Number of Stocks in Your Portfolio Should Be Larger Than You Think: Diversification Evidence from Five Developed Markets. Journal of Investment Strategies 4(1), 43–82. The 73-stock figure is for meeting the standard 90% of the time.
  6. Bessembinder, H. (2018). Do Stocks Outperform Treasury Bills? Journal of Financial Economics 129(3), 440–457. Table 4 reports the share of 20,000 simulated portfolios that beat the value-weighted market; the shares trailing are 100% minus those.
  7. Bessembinder, H. (2026). Returns to “Do-Nothing” Portfolios. SSRN working paper, July 9, 2026 (DOI 10.2139/ssrn.7023959). S&P 500 constituents from CRSP, 1971–2025, before trading costs and taxes. Section II and Table 1 (equal vs. value weighting, including the less frequent rebalancing comparison); Table 6, Panel A (5,000 random draws per start year, share of draws beating the index).
  8. Campbell, J. Y., M. Lettau, B. G. Malkiel, and Y. Xu (2023). Idiosyncratic Equity Risk Two Decades Later. Critical Finance Review 12, 203–223 (NBER Working Paper 29916).
  9. Sharpe, W. F. (1991). The Arithmetic of Active Management. Financial Analysts Journal 47(1), 7–9.
  10. Arnott, R. D., J. Hsu, V. Kalesnik, and P. Tindall (2013). The Surprising Alpha From Malkiel’s Monkey and Upside-Down Strategies. Journal of Portfolio Management 39(4), 91–105. Exhibit 1 (returns, alphas and factor loadings, 1964–2012) and p. 95 (construction; 96 of 100 trials).
  11. Clare, A., N. Motson, and S. Thomas (2013). An Evaluation of Alternative Equity Indices, Part 1: Heuristic and Optimised Weighting Schemes. Cass Business School. Section 7 (10 million random-weight indices, 1969–2011; rolling three-year comparison).
  12. Invesco, Invesco S&P 500 Equal Weight ETF (RSP), read November 18, 2026: 0.20% expense ratio, quarterly rebalancing of the fund and index, inception April 24, 2003, and no capital gains distributions since inception.
  13. State Street Global Advisors, SPY daily holdings file, as of November 17, 2026 (503 equity securities after removing cash and a $0.01 stub line).
  14. Statman, M. (1987). How Many Stocks Make a Diversified Portfolio? Journal of Financial and Quantitative Analysis 22(3), 353–363.
  15. Statman, M. (2004). The Diversification Puzzle. Financial Analysts Journal 60(4), 44–53. Quoted from the abstract.
  16. Vanguard, Vanguard Morningstar Total Stock Market ETF (VTI), read November 18, 2026: 0.03% expense ratio as of April 28, 2026; all 40 distributions from December 2016 to September 2026 were income, with no capital gains.
  17. Chaudhuri, S. E., T. C. Burnham, and A. W. Lo (2020). An Empirical Evaluation of Tax-Loss-Harvesting Alpha. Financial Analysts Journal 76(3), 99–108.
  18. Internal Revenue Service, Publication 550, Investment Income and Expenses (2025), chapter 4: “What Is a Sale or Trade?” and “Wash Sales.”
  19. U.S. Securities and Exchange Commission, Office of Investor Education and Advocacy, Updated Investor Bulletin: Exchange-Traded Funds (ETFs), February 23, 2023.
  20. Internal Revenue Service, Revenue Ruling 2008-5, Internal Revenue Bulletin 2008-3: a loss is disallowed when substantially identical stock is bought in the seller’s IRA.
  21. S&P Dow Jones Indices, S&P U.S. Indices Methodology, November 2026, Eligibility Factors: ineligible structures include limited partnerships and master limited partnerships.
  22. FTSE Russell, Russell US Equity Indexes Construction and Methodology, v7.4, November 2026, section 5.8.1.
  23. Morningstar, Morningstar Market Indexes Methodology Guide, July 2026, chapter 2: limited partnerships are not eligible.

Method

Calibration: monthly total returns (Yahoo Finance adjusted closes) for the 492 SPY constituents with all 60 months of history, November 2021 to October 2026, each regressed on SPY by OLS; company-specific volatility is the annualized residual standard deviation, summarized by its root mean square (30.9%; median 25.3%), which sets the tracking error of an equal-weight portfolio. Simulation: monthly arithmetic returns, index return normal with 9% expected annual return and 15.5% volatility, each stock equal to the index return plus an independent normal shock; 500 companies; seeded. The “measured” variant draws each company’s beta and volatility from the 492 measured pairs with CAPM expected returns (3% risk-free rate); the fat-tail variant uses Student-t shocks with four degrees of freedom. RSP and SPY: month-end adjusted closes, June 2003 to October 2026. The calculator runs the same baseline model in continuous time in your browser. All figures are before trading costs and taxes.

More in Investing & Portfolio

Browse all investing & portfolio guides
Share

Get new guides by email

Evidence-based, no jargon. At most two emails a month. Unsubscribe any time.

Try it in Summitward

See portfolio factor analysis in action with your own financial data. Free to start, no credit card required.

Disclaimer: This tool is for educational and informational purposes only and is not financial, investment, tax, or legal advice. Summitward is not a registered investment adviser, broker-dealer, or financial planner, and no fiduciary relationship is created by your use of it. Consult a qualified professional before acting. Past performance and model projections do not guarantee future results. Provided as is, without warranty of any kind; see our Terms of Service for limitations of liability.