What we tested

Nine trend systems from the arena that are built on moving averages or close relatives: the 50/200 Dual Moving Average, Triple Moving Average, Kaufman Adaptive MA, Hull MA crossover, MACD, MACD with an RSI filter, Ichimoku Cloud, Supertrend, and a dual-MA trend with a macro-event blackout. All run long-only on daily BTC-USD data with walk-forward optimization, before trading costs.

The last five years

Buy-and-hold bitcoin: 7.2% a year, Sharpe ratio 0.33, maximum drawdown −76.6%.

Five systems beat that annual return: Trend + Macro Filter (19.4%), Ichimoku Cloud (11.9%), Dual Moving Average (10.9%), MACD (8.1%) and Supertrend (7.8%). The other four lagged: Triple MA (6.3%), Hull MA (4.3%), MACD + RSI (2.6%) and Kaufman Adaptive MA (2.3%).

The more consistent result is risk. Every system had a smaller maximum drawdown than holding, between −33% (Trend + Macro Filter) and −58% (Hull MA). They achieved it the simple way: they were in the market only 34–53% of the time.

The full history tells a different story

Since our bitcoin data begins in September 2014, buy-and-hold compounded at about 51% a year. Most moving-average systems landed between 43% and 60%: Ichimoku (60.0%) and MACD (58.2%) a bit ahead, Supertrend (42.8%) and the macro-filter version (46.8%) behind, and the MACD + RSI filter far behind (15.5%). Drawdowns were still lower (−46% to −74% vs −83%), but in the explosive early years, being out of the market for half the time was expensive.

Costs matter more than usual

The fast systems trade a lot: the Hull MA crossover made 266 trades in five years and the Kaufman Adaptive MA 187, against 10–12 for the dual and triple MA. Crypto trading costs are higher than for ETFs, so the strategy pages' cost slider is worth a look before drawing conclusions about the fast systems.

Takeaways

  • On bitcoin, simple slow trend filters did their main job in the last five years: similar or better returns with roughly a third less drawdown.
  • They do not reliably beat holding in a strong bull market; their value is surviving the bear markets.
  • Filters on top of filters (MACD + RSI) cut returns more than risk here.
  • Faster systems need a realistic cost assumption.