Tax assumptions: a flat 20% on realized gains, with no state tax. A gain is the sale value minus that position's cost basis. Strategy A sells every Jan 1, even when it buys the same stock back. Strategy B sells all five every Jan 1, and gains and losses net within the year. Net losses carry forward to offset later gains. Tax is paid from the portfolio. Cash from a buyout during the year is taxed in that year. Dividends are left in the total returns and not taxed. At a start after 1999, positions count as bought on Jan 1 of the start year. Lines and stats show positions still open at the end before selling (the S&P 500 is never sold). The "if sold at end" row shows each value after a final sale and tax.
Best-performing S&P 500 stock by calendar year
“Return” = price return in the winning year. “Next year” = total return the following calendar year (what Strategy A earned). flag rows are uncertain, single-source or unverified; hover/tap a status for details. * = partial period to .
Year
Ticker
Company
Return
Next year
Status
Strategy B: each year’s top 5 S&P 500 stocks
Same winner rule (year-end members, full year, price return). “Next yr” = total return of each stock the following year. “Portfolio” = equal-weight (20% each) next-year return, which is what Strategy B earned. Years with incomplete or unverifiable top-5 lists are flagged and not used. Strategy B starts -01-01, the first year after which every top-5 list is complete.
Show top 5 by year ( years)
Year
Top 5 (price return → next-yr total return)
Portfolio next yr
Status
Method
Winner of year Y: highest calendar-year price return among stocks in the S&P 500 at the end of year Y (the convention used by S&P Dow Jones Indices and year-end press lists), requiring a full year of trading. So mid-year additions count (Tesla 2020, Palantir 2024) but SanDisk 2025 (spun off in February) does not.
Strategy A: on Jan 1 of each year put 100% in the previous year’s winner, hold 12 months, rotate.
Strategy B: on Jan 1 put 20% in each of the previous year’s top 5, hold 12 months, and rebalance to equal weight each year. The earliest start with complete, sourced top-5 lists every year is , using the 2008 list (no full top-5 list could be found for 1997, 1998, 2001, 2002 or 2004–2007). On the chart, B starts at the S&P 500’s value on that date. Its stats card uses $100 invested at its own start, and the “since” lines compare all three lines over the same window.
Acquired names inside a holding year (UST 2009, El Paso 2012, old Allergan 2015, Cablevision 2016, Warner Bros. Discovery 2026): valued at the actual deal consideration, with cash held at 0% for the rest of the year. Sources and the arithmetic are in momentum_top5.csv.
Benchmark: S&P 500 Total Return index (^SP500TR, dividends reinvested). Strategy returns are also total returns (dividend-adjusted closes). No costs, taxes or slippage.
Why 1999 and not 1998? Strategy A (1997 winner Dell, +248% in 1998 per S&P DJI) and the S&P 500 could both start in 1998. The page keeps its original 1999 start so earlier results stay comparable.
Window: the backtest starts -01-01, kept from the original version of this page. It runs to the latest close; the last year is year-to-date. The chart uses year-end points, so max drawdown is measured on year-end values; intra-year drawdowns were deeper.
Delisted / acquired names (Dell, Reebok, Providian, Avaya, Family Dollar, XL Capital, Cabot Oil & Gas): for Strategies A and B, none was acquired or delisted inside its holding window. Their returns come from Exa’s per-ticker total-return pages (linked in the data file), with S&P DJI figures for the winning year where available.
Caveats
Hindsight and survivorship: membership comes from a point-in-time constituent list, but Yahoo has no prices for most delisted pre-2010 members, and ticker reuse corrupts some symbols (for example Yahoo’s “EP” is no longer El Paso). Pre-2010 winners therefore rely on published lists, and a few years are flagged.
Sensitive to definitions: with Jan-1 membership instead of year-end, several winners change (2024 would be Vistra, not Palantir; 2020 NVIDIA, not Tesla). Price and total-return rankings can also differ.
Very concentrated: A holds one stock and B five. A’s year-end drawdown reached about −97% in 2000–02. The result depends mostly on the start date. Since 2012 it looks much better than over the full period, which matches what the other 30-year backtest found.