Does the Turtle 20/20/10 breakout strategy on ETH make money?
Short answer: We reproduced the strategy's profit factor and win rate on Binance data; the trade count landed higher because the published snapshot is about a year older. At the advertised 1% of capital per trade it turned 10,000 USDT into about 10,500 in nine years. Long only at 25% per trade, since 2023 it made 13.5% while holding ETH made 124%; its average trade shrank to +0.8% with a 95% lower bound below zero.
The Turtle rules are forty years old: buy a breakout above the 20-bar high, sell when the price falls below the 10-bar low. A public strategy on a popular charting platform applies them to ETH 4-hour candles with one extra filter and reports 281 trades, a 40% win rate, a profit factor of 1.70 and a worst drawdown under 1%. The numbers are stated with costs, which is rare. We asked two things: can we reproduce them, and what does the rule return at a size that matters?
How we tested
- The rule was written and timestamped before any data was pulled.
- Entry when a 4-hour close crosses above the highest high of the previous 20 bars and is also above the highest high of bars 21 to 40 back. Exit when a close falls below the lowest low of the previous 10 bars. Fills at the next open.
- Run 1, the replication: long and short as published, 1% of capital per trade, 0.10% commission and 5 ticks of slippage, all history since the 2017 listing. Target: trade count within 10%, profit factor within 0.2, win rate within 5 points of the published figures.
- Runs 2 and 3: long only, as a spot wallet allows, at 10% and 25% of capital per trade, compounding, 0.15% per side.
- Benchmarks with the same costs: holding ETH, and a daily 200-day EMA filter.
- Verdict at 25% sizing on January 2023 to September 2026, the same window as our earlier 4-hour breakout test. Pass rule: net return above the filter, return over worst drawdown above both benchmarks, and an average trade whose 95% lower bound stays above zero.
Run 1: the replication
| Trades | Profit factor | Win rate | Net | Worst drawdown | |
|---|---|---|---|---|---|
| Published | 281 | 1.70 | 40.2% | +5.2% | 0.95% |
| Ours, to September 2026 | 324 | 1.60 | 38.3% | +5.4% | 1.03% |
| Ours, cut at June 2025 | 283 | 1.61 | 38.2% | +4.8% | 1.03% |
Profit factor and win rate reproduced. The trade count landed 15% high; cut at mid 2025, all three land inside tolerance, so the gap is the age of the published snapshot. The rule was kept as written.
Note what the published figures mean: at 1% of capital per trade, 10,000 USDT became about 10,500 in nine years. The drawdown under 1% is the sizing, since the rule never had more than 1% at risk.
Runs 2 and 3: at a size that matters
January 2023 to September 2026, long only:
| Net | Worst drawdown | Trades | Win rate | Profit factor | Average trade | |
|---|---|---|---|---|---|---|
| Turtle, 10% per trade | +5.7% | 5.4% | 75 | 30.7% | 1.27 | +0.78% |
| Turtle, 25% per trade | +13.5% | 12.9% | 75 | 30.7% | 1.27 | +0.78% |
| Holding ETH | +124.2% | 68.0% | ||||
| 200-day filter | +124.2% | 38.2% | 15 |
Holding and the filter tie on this window to the first decimal by coincidence (124.22% against 124.23%). The Turtle’s average trade of +0.78% on the capital committed has a 95% lower bound of -0.95% over 75 trades, so the data cannot rule out zero. All three parts of the rule fail.
Since 2017, long only at 25%, the rule made +150% against +749% for holding ETH, with a 12.9% drawdown and 7% of capital in the market on average. Its average trade was +2.3% with a lower bound above zero, and nearly all of that came from 2017 to 2022.
What it shows
The published result is honest and small. A positive profit factor on a 1% sizing is a curiosity; scaled to 25% it is a rule that sits in cash 93% of the time and, since 2023, no longer earns its costs with confidence. This matches our earlier test of a 20-bar 4-hour breakout on ETH: short breakouts kept paying fees after 2022 and did not keep up with a one-line daily filter.
Limits
- The charting platform’s feed, bar alignment and tick size differ from Binance API data, which is why the replication tolerance was loose.
- Long only removes the short side that was part of the published number.
- One coin, one rule. Fills at the next open; real fills differ slightly.
Verdict
FAIL. The Turtle 20/20/10 strategy on ETH reproduces as published, and at a usable size it trailed both holding and a 200-day filter since 2023, with an average trade the data cannot separate from zero. Not financial advice.