US stock funds / Rebalancing test
Does rebalancing pairs beat holding?
No for the daily rule after modeled trading costs. Rebalancing only after a larger drift helped slightly, but both rules trailed QQQ in all three tested periods.
A viral Thomas Cover lecture clip describes enormous gains from repeatedly restoring a split between two stocks. We tested a separate, reproducible translation using 13 established US stock funds. This checks whether the idea travels to that fund universe; it does not reproduce the lecture’s chosen stock example.
The drift rule earned 12.8% a year; QQQ earned 19.3%.
That is the latest period, 2020 through July 2026, after modeled trading charges, currency movements and ISK tax. Daily pairs earned 11.2%; holding the same annual fund selections earned 12.7%. The drift rule’s worst fall was also deeper than QQQ’s: about 32% versus 30%.
The funds are historical US ETF proxies. This is not an order plan for a Swedish retail ISK: the exact eligible UCITS products and their tracking differences were not tested. Trading charges include commission and currency conversion; the extra 0.10% per trade is an assumption, not measured execution costs. The tax model applies the 2026 rate throughout history.
01 / What the rule buys and sells
Choose five pairs each year. Then trade daily or after a five-point drift.
At each year-end, rank all pairs by how differently their daily returns moved over the previous 252 trading sessions. Take five pairs without reusing a fund, put an equal amount into each pair and aim for half in each partner. That makes one portfolio of ten funds. The result below belongs to that portfolio, rather than an average of separately tested pair strategies.
Daily pairs restore the split every trading session. Low-correlation drift pairs act when a partner’s share of its pair is at least five percentage points away from 50%. Both fix whole-share orders at that day’s close and execute at the following close, so actual fill prices can differ from the signal price.
Hold the same funds uses identical annual selections, starting pair weights and annual replacement dates, and skips the intervening rebalances. It still makes necessary trades to fund tax. Hold QQQ and Hold SPY retain those funds, reinvest available distributions monthly and fund the same modeled tax.
02 / Returns after trading costs
Neither pair rule caught QQQ—even with trading charges removed.
The drift rule added roughly 0.1 percentage point a year over holding the same funds in each period after costs. Daily trading cost more than its rebalancing benefit. Use the controls to compare all three periods and remove or increase the assumed charges.
Each period starts with SEK 1 million. Currency movements and modeled ISK tax stay in every view. Ending balances include a final sale.
| Rule | Per year | Worst fall | Ending SEK |
|---|---|---|---|
| Hold QQQ | 19.3% | 29.7% | 3,189,000 |
| Hold SPY | 14.4% | 30.3% | 2,424,000 |
| Low-correlation drift pairs | 12.8% | 32.5% | 2,206,000 |
| Hold the same funds | 12.7% | 32.2% | 2,193,000 |
| Daily pairs | 11.2% | 32.9% | 2,016,000 |
“Per year” is the constant growth rate equivalent to the whole period’s gain. “Worst fall” is the largest daily drop from an earlier account peak. Each period follows real history in order; these are three separately funded accounts.
03 / Why the portfolios separated
Different fund exposure helped in 2022, but missed much of QQQ’s rallies.
The pair portfolio spread money across ten funds and repeatedly moved money from relative winners to their partners. That is a different investment from keeping everything in QQQ. In the SEK account, drift pairs gained about 5% in 2022 while QQQ lost 23%; in 2020 the pairs lost 3% while QQQ gained 26%, and in 2023 they gained 12% while QQQ gained 48%.
Holding the same selected funds followed almost the same path as drift pairs. The choice of funds explains most of the gap with QQQ; the extra rebalancing trades explain very little. Daily trading then reduced what the investor kept.
Check the rules, assumptions and sensitivity results
The fixed universe contains SPY, QQQ, IWM, MDY and the nine original Select Sector SPDR funds: XLB, XLE, XLF, XLI, XLK, XLP, XLU, XLV and XLY. All existed before the first selection at the end of 2004. These stock funds overlap; this does not test stock–bond or other mixed-asset rebalancing.
Pairs are chosen greedily from the lowest preceding-year correlations, breaking ties by ticker. Each fund can appear only once. The strategy keeps the same lookback, five pairs, 50/50 target and five-point band in all periods. It was not tuned to these results. Daily timing was added before any ETF returns were inspected, after retrieving the original post’s explicit daily claim.
All accounts begin with SEK 1 million. The model uses whole shares, a 1% purchase reserve, next-close execution and a per-order ceiling of 0.1% of the previous 60 sessions’ average dollar volume. Trading commission is 0.25%, minimum USD 1; currency conversion is 0.25% per trade and modeled distribution conversion. The base execution haircut is 0.10% per side and the higher-cost check uses 0.25%. “No trading charges” removes those charges, while retaining currency movements and tax.
ISK uses a constant annual 1.065% tax rate on quarterly account values, with the tax-free allowance assumed used elsewhere. Prior-year tax is funded the following May. This applies 2026 account conditions to old market history; it is not a reconstruction of the actual tax rates or availability of ISK in every historical year. US distribution withholding is 15%, with a bounded tax credit.
The price source gives distribution amounts but not broker payment dates. The main model pays their cash equivalent 20 trading sessions after the ex-date; repeats use zero and 40 sessions. These timing changes leave daily pairs behind holding the same funds, and both pair rules behind QQQ in every period. Cash-equivalent treatment of any noncash fund distribution and the exact broker payment date remain modeling limits.
The same comparison in dollars, without currency movements, tax or trading charges, also leaves both pair rules behind QQQ in all three periods:
| Period | Daily pairs | Drift pairs | Same funds | QQQ |
|---|---|---|---|---|
| 2005–2013 | 8.4% | 8.3% | 8.2% | 10.0% |
| 2014–2019 | 11.0% | 10.8% | 10.7% | 17.2% |
| 2020–July 2026 | 13.7% | 13.5% | 13.4% | 19.9% |
The 120 account cases combine five policies, three periods, currencies, costs and distribution timing. They are variations on one real market history, not 120 independent pieces of evidence. Every saved account path was rebuilt by a separate accounting implementation; all 22 annual pair selections and all 120 terminal sales were independently checked. Future-price mutations left earlier decisions unchanged.
The lecture’s chosen stock example is not reproduced here. This fund result does not refute Cover’s mathematical theorem or establish how frequently other rebalancing strategies succeed. It tests these fixed rules in this named fund universe.
04 / Read the original claim
A successful example still needs a rule for choosing tomorrow’s investments.
The hard question is which investments to own before their future returns are known. Lower past correlation did not solve that problem in this fund test. Rebalancing can still help maintain a chosen risk allocation; that is a different objective from beating QQQ.
Original X post · Full Thomas Cover lecture · Trading-charge schedule · Currency-conversion charges · Skatteverket: ISK
Published 14 September 2026. Market history ends 31 July 2026. Returns are historical simulations, not a forecast.
Use this result
See how much you need to retire—and compare what to invest in.
Enter what you have and what you want to spend each month. Compare three portfolios against the same market history. Your financial figures stay in your browser.
Build my free FIRE plan