Trend-Following Backtester · Guide · backtest 한국어

equity · JP

Metaplanet Inc. backtest

10 trend-following strategies were compared on the full daily history of Metaplanet Inc.. 10 beat buy-and-hold CAGR; Donchian 채널 돌파 ranks first by Sharpe, matching the main backtester.

Data 2004-11-16 ~ 2026-07-31daily bars 5,362 (21.7 years)Costs 0.5%Venue TSEMonthly snapshot 2026-07-31

Results by strategy

10 of 10 beat buy and hold (CAGR -18.4%, drawdown -100.0%).

Sharpe leader (same as main): Donchian 채널 돌파 — CAGR 20.4%, MDD -91.3%, Sharpe 0.60, 28% exposure.

No strategy retained at least 75% of buy-and-hold CAGR while reducing drawdown by both 20% relatively and 5 percentage points absolutely.

The alternative must retain 75% of buy-and-hold CAGR, improve drawdown by at least 20% and 5 percentage points, and stay below 95% exposure; the ranking weights return retention 60% and drawdown improvement 40%.

Strategy · parametersCAGRMDDSharpeCalmarTotalTradesExposure
Donchian 채널 돌파
entryN=75, exitN=52
20.4%-91.3%0.600.2256.2x2228%
볼린저 밴드 돌파
n=46, k=2.5
13.1%-97.3%0.480.1314.4x7018%
Keltner 채널 돌파
emaPeriod=40, atrPeriod=16, mult=3.6
12.0%-93.1%0.470.1311.8x3811%
SMA 크로스오버
fast=9, slow=132
10.8%-94.0%0.470.119.2x4630%
ADX / DI 방향성
period=31, threshold=14
9.4%-93.1%0.460.107.0x9841%
EMA 크로스오버
fast=13, slow=96
9.0%-97.2%0.440.096.5x7228%
ROC 모멘텀
n=32, threshold=0.11
9.2%-97.4%0.430.096.8x25221%
Supertrend
period=14, mult=4.4
7.8%-98.4%0.430.085.1x7040%
Parabolic SAR
step=0.025, maxStep=0.22
5.1%-98.5%0.400.053.0x39240%
MACD
fast=15, slow=33, signal=15
-8.8%-98.0%0.23-0.090.1x29552%
Buy and hold-18.4%-100.0%0.25-0.180.0x1100%
0.001 0.01 0.1 1x 10x 100x 2005 2008 2011 2014 2017 2020 2023 2026 Donchian 채널 돌파Buy and hold
Donchian 채널 돌파: Solid strategy line versus dashed buy-and-hold line. Log scale.
0% -25% -50% -75% -100% 2005 2008 2011 2014 2017 2020 2023 2026 Buy and holdDonchian 채널 돌파
Donchian 채널 돌파 maximum drawdown -91.3% (2013-02-27 → 2022-03-09), recovered after 4,164 days.

Rules of the top strategy

A new high means a trend starting; a new low means it’s over (turtle-trading style).

Method and caveats

Past performance does not guarantee future returns. Parameters were selected on the same full-history sample using sharpe over 40 trials per strategy, so results contain in-sample bias. 0.5% trading costs are included; taxes, slippage and rejected orders are not.
Re-run the same setupAll asset reports

Read next