−1.50+1.5
Sharpe ratio +0.469 — return earned per unit of risk taken; 0 means no edge, 1 is strong. The 95% interval is -0.22 to +1.16. It crosses zero — not statistically significant. Suggestive, not proven — the reason to trust it is that trend-following is a documented effect across markets, not that this backtest is decisive.
Worst drawdown, this rule-74.6%Worst drawdown, buy & hold-89.0%Time in the market47%
Drawdown is the deepest fall from a previous peak — the most you would have been down had you started at the worst possible moment.
What the trading costs. Priced at 55 basis points per side — Kraken's taker fee plus the spread you actually cross. This is what you pay, not a wholesale rate.
Round trips per year11Cost of one round trip1.10%Fees burned per year12.1%
The 50-day average flips more than once a month, and over half of those states last five days or less — the rule changes its mind and changes it back. Every flip is a full sale and repurchase of all twenty coins. That is where 12.1% of the portfolio a year goes.
The window is a guess. Sharpe by window:
50-day window+0.469100-day window+0.337200-day window+0.198
Picking the best of these after seeing the table is how backtests lie. That spread is why the page shows all three calls rather than one.
Lately. The numbers above lean on the 2020–21 bull market. This is the last 2.7 years, held back and never used to choose the rule or its window. Since 2024 this rule has lost money — and lost more than doing nothing. It is down 17.6% a year after costs, against 11.5% a year for simply holding the basket: it is 6.1% a year behind the do-nothing option. The reason is on this page — it trades 14 round trips a year and burns 15.8% of the portfolio annually in fees to do it. Its one remaining defence is a shallower fall: -71.1% at worst versus -75.8% for holding. That is what you are buying, and you are paying return for it.
Known weakness: an average is slow by construction, so this rule lags a sharp V-shaped recovery and buys back in late.