Buying a new 20-day high is one of the oldest trend-following rules. The Turtles, the traders Richard Dennis trained in the 1980s, used it as the entry of their System 1. Bitcoin trends harder than almost anything, so if the rule works anywhere, it should work here. It is also the opposite of the Lab’s RSI test on Bitcoin: that one buys weakness, this one buys strength.
I tested a simple, close-based version on Bitcoin’s daily bars since 2014: buy a close above the 20-day high, sell a close below the 10-day low. 63 closed trades, and one still open.
The rules
- Buy when Bitcoin closes above the highest high of the previous 20 days.
- Sell when it closes below the lowest low of the previous 10 days.
- A simplified, close-based version of the Turtles’ System 1: the same 20-day entry and 10-day exit, without their skip filter, 2N stop or adding to the position. Daily bars. Long only; in cash between trades.
Market: Bitcoin in US dollars (BTC-USD), trades every day.
The result
| 20-day breakout | Buy-and-hold | |
|---|---|---|
| Annual return | 53.90% | 58.68% |
| Total return | 17539% | 25346% |
| Max drawdown | −56.7% | −83.4% |
| Worst fall, from – to | Mar 2021 – Dec 2022 | Dec 2017 – Dec 2018 |
| Time in the market | 46% | 100% |
| Trades | Value |
|---|---|
| Closed trades | 63 (+1 open) |
| Winners | 28 (44.4%) |
| Average winner | +39.69% |
| Average loser | −6.61% |
| Win/loss ratio | 6.01 |
| Profit factor | 4.81 |
| Best / worst trade | +354.2% / −21.8% |
| Average holding time | 32 days |
My read
Most trades lost: only 44% made money. That is normal for trend following, and the reason it still works is size. The average winner made 39.7%, the average loser lost 6.6%, about 6 times less. Losers were out in 16 days on average; winners ran for 52.
2017 and 2018 show how it feels. It sold the December 2017 top within days, on Dec 23, 2017, with a 90.7% gain on that trade. Then it lost 6 times in a row, from Feb 2018 to Mar 2019, each loss cut before it got big. The Mar 2019 breakout ended the streak with a 94.2% gain. Few people would still be following the rule by then, and that one trade more than made back all of those losses.
The best trade bought in Oct 2020 and sold in Mar 2021 for a 354% gain. Chain all the closed trades together and $1 becomes about $178; take out that one trade and it is about $39. Trend following lives on a few outliers, and you only get them if you take every signal, because nobody knows in advance which one it is.
Over the whole period it made 53.9% a year, while holding Bitcoin made 58.7% a year, so it did not beat simply holding. What it bought was a smaller worst fall: 57% against 83%. Its worst stretch ran from Mar 2021 to Dec 2022, and it holds the worst single trade: bought in Sep 2022, sold in the FTX collapse of Nov 2022 for a 21.8% loss. All of this is before costs, at about 5 round trips a year.
It is in a trade right now: bought on Sep 22, 2026, 0.7% down on Oct 6, 2026.
My read: this is the first rule in the Lab that behaves the way its fans say it does. Many small losses, a few huge winners, and the big ones pay for everything. It still did not beat holding, and a 57% fall is more than most people can sit through. For me the 20-day breakout is the skeleton of a trend system, not the whole system. The Turtles put a stop and position sizing on top of it, and those are the parts I would not trade without.
Every trade
Open at the end of the data: bought Sep 22, 2026 at 86,172, −0.7% at the last close (Oct 6, 2026). Not counted in the trade statistics.
63 closed trades, newest first
| Bought | Price | Sold | Price | Result |
|---|---|---|---|---|
| 2026-07-15 | 64,712 | 2026-09-16 | 76,150 | +17.7% |
| 2026-04-14 | 74,182 | 2026-05-17 | 77,429 | +4.4% |
| 2026-03-05 | 70,841 | 2026-03-22 | 67,845 | −4.2% |
| 2026-01-05 | 93,883 | 2026-01-21 | 89,377 | −4.8% |
| 2025-10-02 | 120,681 | 2025-11-05 | 103,892 | −13.9% |
| 2025-09-17 | 116,469 | 2025-09-23 | 112,015 | −3.8% |
| 2025-08-14 | 118,360 | 2025-08-20 | 114,275 | −3.5% |
| 2025-07-10 | 115,987 | 2025-08-02 | 112,527 | −3.0% |
| 2025-04-23 | 93,699 | 2025-05-31 | 104,638 | +11.7% |
| 2024-10-16 | 67,613 | 2025-02-03 | 101,405 | +50.0% |
| 2024-09-19 | 62,940 | 2024-10-02 | 60,633 | −3.7% |
| 2024-08-24 | 64,179 | 2024-09-02 | 59,112 | −7.9% |
| 2024-07-16 | 65,097 | 2024-08-03 | 60,680 | −6.8% |
| 2024-05-16 | 65,232 | 2024-06-15 | 66,191 | +1.5% |
| 2024-02-08 | 45,302 | 2024-03-20 | 67,914 | +49.9% |
| 2023-10-17 | 28,416 | 2024-01-15 | 42,512 | +49.6% |
| 2023-10-02 | 27,531 | 2023-10-12 | 26,757 | −2.8% |
| 2023-06-21 | 30,027 | 2023-07-25 | 29,227 | −2.7% |
| 2023-03-15 | 24,376 | 2023-04-22 | 27,818 | +14.1% |
| 2023-02-16 | 23,623 | 2023-03-04 | 22,353 | −5.4% |
| 2023-01-09 | 17,197 | 2023-02-10 | 21,651 | +25.9% |
| 2022-12-14 | 17,816 | 2022-12-17 | 16,795 | −5.7% |
| 2022-09-13 | 20,297 | 2022-11-09 | 15,881 | −21.8% |
| 2022-07-19 | 23,389 | 2022-08-20 | 21,166 | −9.5% |
| 2022-03-25 | 44,349 | 2022-04-07 | 43,504 | −1.9% |
| 2022-02-08 | 44,118 | 2022-02-18 | 40,031 | −9.3% |
| 2021-11-09 | 66,972 | 2021-11-17 | 60,368 | −9.9% |
| 2021-10-06 | 55,361 | 2021-10-28 | 60,622 | +9.5% |
| 2021-07-26 | 37,338 | 2021-09-21 | 40,694 | +9.0% |
| 2021-04-14 | 63,110 | 2021-04-23 | 51,094 | −19.0% |
| 2020-10-11 | 11,384 | 2021-03-25 | 51,704 | +354.2% |
| 2020-07-23 | 9,581 | 2020-08-26 | 11,488 | +19.9% |
| 2020-06-02 | 9,530 | 2020-06-12 | 9,481 | −0.5% |
| 2020-04-24 | 7,551 | 2020-05-25 | 8,907 | +18.0% |
| 2020-04-07 | 7,176 | 2020-04-16 | 7,117 | −0.8% |
| 2020-01-07 | 8,164 | 2020-02-26 | 8,821 | +8.0% |
| 2019-10-27 | 9,552 | 2019-11-09 | 8,814 | −7.7% |
| 2019-08-06 | 11,478 | 2019-08-15 | 10,312 | −10.2% |
| 2019-06-21 | 10,145 | 2019-07-15 | 10,895 | +7.4% |
| 2019-03-21 | 4,029 | 2019-06-05 | 7,824 | +94.2% |
| 2019-02-19 | 3,947 | 2019-03-05 | 3,896 | −1.3% |
| 2018-08-29 | 7,047 | 2018-09-07 | 6,467 | −8.2% |
| 2018-07-18 | 7,371 | 2018-08-04 | 7,033 | −4.6% |
| 2018-07-08 | 6,774 | 2018-07-13 | 6,238 | −7.9% |
| 2018-04-21 | 8,896 | 2018-05-12 | 8,505 | −4.4% |
| 2018-02-21 | 10,690 | 2018-03-09 | 9,338 | −12.7% |
| 2017-11-17 | 7,709 | 2017-12-23 | 14,699 | +90.7% |
| 2017-10-02 | 4,409 | 2017-11-12 | 5,950 | +34.9% |
| 2017-08-06 | 3,214 | 2017-09-14 | 3,155 | −1.8% |
| 2017-04-07 | 1,177 | 2017-07-11 | 2,338 | +98.6% |
| 2017-02-01 | 989 | 2017-03-11 | 1,176 | +18.9% |
| 2016-10-08 | 619 | 2017-01-07 | 909 | +46.8% |
| 2016-09-04 | 609 | 2016-09-22 | 596 | −2.0% |
| 2016-05-28 | 530 | 2016-06-23 | 624 | +17.7% |
| 2016-03-28 | 424 | 2016-05-20 | 443 | +4.3% |
| 2016-02-15 | 400 | 2016-03-05 | 401 | +0.1% |
| 2015-12-06 | 389 | 2015-12-27 | 423 | +8.8% |
| 2015-10-07 | 243 | 2015-11-12 | 338 | +39.2% |
| 2015-09-09 | 238 | 2015-09-22 | 231 | −3.2% |
| 2015-06-17 | 249 | 2015-08-09 | 265 | +6.3% |
| 2015-05-04 | 239 | 2015-05-19 | 232 | −3.0% |
| 2015-03-03 | 282 | 2015-03-19 | 261 | −7.4% |
| 2014-11-13 | 421 | 2014-11-22 | 353 | −16.1% |
How it was tested
- Signals are read on the daily close; each trade happens at the next day’s close. No look-ahead.
- Long only, fully in or fully out. No commissions, slippage, interest or dividends.
- The test starts on the first day every indicator has a value (Oct 7, 2014); buy-and-hold covers the same days.
- Prices: Yahoo Finance daily data (BTC-USD), snapshot
btc-2026-10-07.csv. Checked by a second, independent program before publishing.
Not financial advice. This is educational content, shared for information and entertainment only, and it is not a recommendation to buy or sell any asset. Backtests and past results do not guarantee future results, and the numbers or charts here may contain mistakes. Trading carries risk, including losing all the money you put in. Do your own research before you make any decision.