The Lab

How we tested

• Prices: Bitcoin's weekly closing price, 2017 to September 2026. Futures funding from 2020.
• The plan, each week:
  • Sell a slice (10%, 20% or 30% of the Bitcoin held) when the price is 10%, 20% or 40% above its 20-week average.
  • Buy back with that slice's cash once the price falls to the sale price, or 5% or 10% below it. There is no time limit, so it waits for as long as it takes.
  • One sale at a time, or a ladder: more sales while one is waiting, each only at a higher price than the last.
  • That makes 54 versions. Fees of 0.1% each way. Waiting cash earns nothing.
• Fair judging:
  • The version with the most Bitcoin in 2018 to April 2022 was picked first, then judged on May 2022 to September 2026. Each half started from 1 BTC.
  • It was compared with the same number of sales on random weeks (1,000 tries), and on evenly spaced weeks.
  • To pass, it had to beat holding in both halves and beat at least 95% of the random tries.
• The hedge: short 20% or 30% of the Bitcoin on futures, reset each week, with the gains and losses turned back into Bitcoin.

The maths, step by step

1. One round that works. Sell 0.1 BTC at 100,000 for 10,000. The price falls to 90,000. Buy back:

10,000 ÷ 90,000 = 0.111 BTC. You now own 0.011 BTC more than before.

2. One round that doesn't come back. In December 2023 the plan sold 0.131 BTC at 43,790, for about 5,720. Bitcoin never came back below the buy-back price. At September 2026's price, that cash buys about 0.068 BTC. If it never refills, the sale cost 0.063 BTC.

3. Why the hedge loses Bitcoin when the price rises. Short 0.2 BTC at 100,000, and the price rises to 125,000.

• The short loses 0.2 × 25,000 = 5,000, which is 0.04 BTC at the new price.
• A week of funding at a typical 0.2% earns 0.2 × 100,000 × 0.2% = 40, about 0.0003 BTC.
• One week's rise of that size wipes out years of funding.

What we found

1. Most versions end with less Bitcoin.

• Only 2 of the 54 beat holding in 2018–2022, and 15 in 2022–2026.
• The typical version ended with about 0.85 to 0.88 BTC for each 1 BTC held.
• Selling big slices often was the worst. 30% slices at just 10% above average ended with 0.24 to 0.30 BTC.

2. The best version: sell 10% when the price is 40% above its 20-week average, buy back 10% lower, as a ladder.

• It beat holding in both halves: +2.7% and +3.1% more Bitcoin.
• But the same number of sales on random weeks did almost as well as often. It beat only 87% of them, against the 95% needed. It does not pass.
• Selling only when far above average did help, compared with selling at other times: random weeks gave 0.945 BTC, evenly spaced weeks 0.877. It just wasn't enough to be sure.

3. It depends on when you look.

• Run from 2018 to 2026, the best version ended with 1.35 BTC for each 1 BTC held. It was ahead at 86% of week ends.
• But in the middle of the 2021 bull run it was 21% behind (0.79 BTC): its sales were waiting for prices that hadn't come back yet.
• Over any 2 years it was ahead 62% of the time, and over any 4 years 94% of the time. Short horizons are a coin flip; long ones favour it a little.

4. Sales can wait a long time.

• Sales at 18,000–19,000 in late 2020 waited about two years to buy back, until the November 2022 low.
• One sale is still waiting: sold at 43,790 in December 2023.

5. The hedge loses Bitcoin in a rising market.

• From 2020, a 20% short lost 0.27 BTC to the price rising and earned 0.13 BTC in funding. It ended at 0.86 BTC. A 30% short ended at 0.76.
• Funding never made up for the price rise, counted in Bitcoin. The hedge keeps dollars steady, not Bitcoin.

What it means

• Holding is a hard benchmark. Bitcoin rose over these years, so every sale is a small bet that the price comes back. Usually it does, but late, and sometimes not at all.
• The least-bad version sold small slices (10%), only when the price was far above its average (40% above the 20-week average), and bought back a little lower. Even that spent long stretches behind during bull runs, and it didn't clear the bar against random sales.
• This is a one-sided grid. It sells first and buys back later. It is also a one-sided cousin of "Shannon's demon", the rebalancing idea from the mathematician Claude Shannon. Like our earlier grid and rebalancing studies, and an earlier test of a buy-low, sell-in-quarters ladder (which lost to holding in 52 of 54 rising-market runs), it tends to sell too early in a rising coin.
• Hedging doesn't grow your Bitcoin. It earns dollars, not coins.
• Waiting cash earning 4% a year would add a little (+3.1% becomes +4.2% more Bitcoin). Not enough to change the verdict.