Rebalancing premium in crypto: a 7-coin backtest, 2020–2026

Rebalancing means selling what has risen and buying what has fallen to get the weights back to target. The idea that this alone earns a return, the "rebalancing premium" or diversification return, goes back to Scott Willenbrock's 2011 paper in the Financial Analysts Journal¹. Daniela Hanicová and Radovan Vojtko tested it on cryptocurrencies in a study Quantpedia summarized in December 2021²: 27 coins from 2018-12-31 to 2021-10-29, an equal-weight basket rebalanced every day against the same basket bought once and never touched. They found a premium, and they report that daily rebalancing did better than monthly.

This article repeats the test on Binance spot from 2020 to 2026, with trading costs, several rebalancing schedules, and two coins that died along the way.

Why rebalancing can earn anything

Take two coins that both end where they started. Coin A doubles and then halves; coin B halves and then doubles. Hold both without touching them and the basket ends flat. Rebalance to 50/50 after the first step: the basket is worth 1.25 of its start (half of it doubled, half halved), it is split 0.625/0.625, and after the second step it is worth 0.625 × 0.5 + 0.625 × 2 = 1.5625 — up 56.25% on two coins that made nothing. Rebalancing sold A after it rose and bought B after it fell, and both moves reversed.

Two things follow. The premium grows with volatility and with how independently the coins move. And it needs moves that reverse: a coin that keeps falling to zero is bought again at every rebalance.

Data and method

• Coins: BTC, ETH, BNB, XRP, ADA, SOL and DOGE against USDT, Binance spot, daily candles.
• Window: 2020-09-01 to 2026-10-21. It starts after SOL's spot listing (its candles begin on 2020-08-11), so all seven coins are in the basket from the first day.
• One backtest per coin: buy once at the next open with 9,980 USDT of a 10,000 USDT account and hold to the end of the window. 0.1% fee and 0.05% slippage per side. The runs are #1496–#1502, listed in the table below.
• Baskets, built from each run's daily equity series — the series a report's Sharpe ratio comes from:
  • never rebalanced: equal money in each coin on 2020-09-01, then nothing;
  • rebalanced: back to 1/7 per coin at the day's close on a schedule — daily, weekly (Sunday), month-end, quarter-end, year-end — or whenever a weight leaves a band of ±25% around its target (10.71%–17.86%);
  • every rebalance pays 0.15% of the amount traded, the same fee plus slippage as the runs.
• Cross-check: on Janorium you can select finished backtests, choose "Combine into portfolio" and get the Portfolio preview. It combines the runs' daily returns at fixed weights, which is a basket rebalanced every day with no trading costs. For these seven runs it reports +3,074.26%, the same as this calculation without costs.

The never-rebalanced basket is simply the average of the seven runs: +1,307.23%.

Backtest

Buy & hold sleeve (spot 1d)

BTCUSDT · 1d · Spot · 2020-09-01 → 2026-10-22

Return

+602.95%

Max drawdown

76.62%

Sharpe

0.84

Win rate

100.0%

Trades

1

Open the report →

Every schedule beat holding

Turnover is how much of the basket was bought in a year (sales are the same amount); costs are their yearly average as a share of the basket. Drawdowns here are measured on daily closes and read shallower than an intraday measure would.

Every rebalanced basket in the table ended more than twice as high as the untouched one. Trading costs did not decide anything: even daily rebalancing paid under 1% a year. The untouched basket drifted into three coins — at the end BNB was 30.56% of it, DOGE 25.59% and SOL 24.94%, ADA 1.83% — and fell further in the bad years: 85.65% from its peak against about 79% for the rebalanced ones.

The same test on six coins from 2020-01-01 (without SOL, runs #1503–#1508) gives the same picture: +2,213.39% never rebalanced, +5,143.57% daily, +6,908.11% at month-end.

Backtest

Buy & hold sleeve (spot 1d)

DOGEUSDT · 1d · Spot · 2020-09-01 → 2026-10-22

Return

+2,420.33%

Max drawdown

92.33%

Sharpe

0.79

Win rate

100.0%

Trades

1

Open the report →

The date matters more than the frequency

The original study found that daily rebalancing beat monthly. Here month-end beat daily — but month-end is just one of 28 days a monthly rebalance could fall on. Rebalancing on every possible weekday, day of the month and month of the year shows how much of the gap is the calendar:

Daily rebalancing made +2,891.89%; never rebalancing, +1,307.23%.

• Every schedule on every date beat holding. The worst case, a yearly rebalance at the wrong month-end, still made +2,235.27%.
• Daily came in below the median weekly, monthly and quarterly result — here, in the six-coin basket from 2020 and in the seven coins from 2022-04-16. Before costs too: +3,074.26% against a monthly median of +3,359.01%.
• But moving a monthly rebalance by a few days moved the result by more than switching from daily to monthly. The same monthly rule made anywhere from +2,546.30% to +4,821.62% depending on the day. Neither "daily is better" nor "monthly is better" holds up as a rule on this data.

Within one year, holding usually won

The premium did not show up year by year. A basket formed fresh on each 1 January:

Rebalancing lost in four of the five full years. The same holds for rolling windows starting on the first of every month from September 2020: over one year, daily rebalancing beat holding in 24 of 62 windows and month-end in 28. Over two years they won 43 and 50 of 50, over three years 33 and 37 of 38, over four years all 26.

Those longer windows overlap and almost all of them contain the 2021 boom and the 2022 crash, so they are close to one observation, not dozens. One reading: within a year the leaders kept leading, and selling them early cost money; across years the leaders of 2021 gave much of it back, and the rebalanced basket had already sold them.

The premium also leans on one coin. Portfolio preview puts DOGE's volatility at 199.97% a year, from daily returns, against bitcoin's 57.04%. Without DOGE the daily rule's lead over holding shrinks from 20.09 to 8.32 percentage points of CAGR (month-end: from 27.03 to 17.61); without any other single coin it stays between 15.98 and 25.50.

When a coin dies

The seven coins above are survivors: they are big today, which is why they were picked. Rebalancing has a specific risk with coins that do not survive — it keeps buying them on the way down. Two coins that collapsed are on Janorium only as delisted Binance USDT-M perpetuals with hourly candles: LUNA from 2021-01-28 to 2022-05-12 and FTT from 2022-04-15 to 2022-11-14. They were held the same way through the perpetual: 1x, isolated margin, funding switched off so the run follows the price as a spot holder would see it, same fees. When a coin's candles end, its remainder is sold at the last close; the rebalanced baskets spread the money over the other coins, the untouched basket keeps it as cash.

Backtest

Buy & hold sleeve (perp 1h, 1x, no funding)

LUNAUSDT · 1h · Futures · 2021-02-01 → 2022-05-13

Return

-99.53%

Max drawdown

99.99%

Sharpe

0.94

Win rate

0.0%

Trades

1

Open the report →

Backtest

Buy & hold sleeve (perp 1h, 1x, no funding)

FTTUSDT · 1h · Futures · 2022-04-16 → 2022-11-15

Return

-96.25%

Max drawdown

97.14%

Sharpe

-1.39

Win rate

0.0%

Trades

1

Open the report →

From 2021-02-01, with LUNA as an eighth coin, rebalancing gained from it. LUNA's run grew about 64 times from the start of this window to its peak on 2022-04-04, and the rebalanced baskets sold it all the way up.

The quarter-end result is the calendar again: its rebalance on 2022-03-31 sold LUNA four days before the top.

From 2022-04-16, with LUNA and FTT as the eighth and ninth coins, daily rebalancing lost to holding. This window starts where FTT's candles begin, with LUNA close to its peak.

Without the two coins, daily rebalancing was 17.47 percentage points ahead of holding; with them it was 10.20 behind. LUNA's last four days were -53.13%, -41.85%, -93.77% and -99.38%. Every evening the daily rule topped LUNA back up to a ninth of the basket, so it lost that ninth again and again; the untouched basket lost its LUNA only once. FTT did the same in November 2022: -75.09% and -67.00% on the 8th and 9th, then +94.78% on the 10th. The month-end basket had last rebalanced on 2022-04-30, so it too lost its LUNA only once — and still beat holding.

Conclusion

• An equal-weight basket of seven large coins rebalanced on any schedule beat the same basket left alone from 2020 to 2026: +2,235.27% to +4,821.62% against +1,307.23%, with smaller drawdowns, after trading costs.
• The premium showed only over two years or more. Within single years, holding usually won.
• How often to rebalance mattered less than which day: the same monthly rule ranged from +2,546.30% to +4,821.62%. Daily rebalancing, the study's choice, sat below the typical weekly, monthly and quarterly result.
• With coins that collapse, daily rebalancing buys the collapse every day. Adding LUNA and FTT from April 2022 turned it from 17.47 points ahead of holding to 10.20 behind, while a monthly rebalance still came out ahead.

How to put such a test together on the platform: how to backtest a crypto strategy. For buying in one go versus in parts, see DCA or lump sum.

Notes

1. Scott Willenbrock, "Diversification Return, Portfolio Rebalancing, and the Commodity Return Puzzle", Financial Analysts Journal 67(4), 2011; working paper at arxiv.org/abs/1109.1256. Checked 2026-10-22. ↩

2. Daniela Hanicová and Radovan Vojtko, "Rebalancing Premium in Cryptocurrencies", SSRN abstract 3982120; the 27 coins, the period and the daily-versus-monthly comparison are from the Quantpedia summary of 2021-12-13. Checked 2026-10-22. ↩