One strategy, one account
Each strategy is measured on its own, as if it had its own account. Strategies are not combined, and capital they would compete for in a shared account is ignored.
Fixed capital, not compounded
A month’s return is the profit or loss of the trades closed that month, divided by the capital at the start. Profits are treated as withdrawn and losses as topped up, so a period’s return is the sum of its monthly returns. Position sizes are the ones each strategy uses live.
Backtest, then live
Months before 21 September 2026 come from historical simulations of the current rules, using historical exchange price data and fees. From 21 September 2026 the live strategies use the trading bot’s own closed trades, net of fees and funding, divided by account equity at entry. September 2026 mixes both. Trend Following is a backtest throughout; its live version is still paper trading.
What backtests miss
- The rules were designed on much of the same history, so backtests are in-sample and likely overstate future returns.
- Fills in thin markets are modelled, not observed. Several strategies only work with small capital.
- Month-end numbers hide intra-month drawdowns, which can be far deeper.
- Patterns can weaken or stop working, and some strategies pause themselves when that happens.
Metric definitions
All metrics use the monthly returns shown on each strategy page (r₁ … rₙ, usually n = 24) with a risk-free rate of 0%. Because capital is fixed, months add rather than compound.
- Fixed capital
- Monthly return = closed-trade profit or loss / starting capital. Period returns add, rather than compound; profits are treated as withdrawn and losses as topped up.
- Annualised return
- Mean monthly return × 12, including zero-return months; expressed as a percentage of fixed capital, not a compound growth rate.
- Annualised volatility
- Sample standard deviation of monthly returns × √12, including zero-return months; expressed as a percentage.
- Sharpe ratio
- Annualised return / annualised volatility, with a risk-free rate of 0%. Undefined for zero volatility; short samples have wide uncertainty.
- Sortino ratio
- Annualised return / (√(mean of min(r, 0)²) × √12), using all monthly observations. Undefined without losing months.
- Max drawdown
- Largest fall in the arithmetic cumulative monthly return from a prior month-end peak, measured as a percentage of fixed starting capital. It does not measure losses within a month.
- Calmar ratio
- Annualised return / absolute maximum month-end drawdown. Undefined without a drawdown.
- Longest drawdown
- Maximum consecutive month-ends below a previous cumulative-return peak, measured in months.
- Positive months
- Positive-return months / non-zero-return months. Zero-return months are excluded from this share; a zero return does not establish that a strategy was out of the market.
- Profit factor
- Sum of positive monthly returns / absolute sum of negative monthly returns. This is a monthly aggregation, not a trade-level profit factor.
- Correlation and beta to BTC
- Pearson correlation of paired monthly strategy and Bitcoin benchmark returns; beta = sample covariance / sample variance of Bitcoin returns. Requires at least 12 paired months and non-zero variance.
- n/m
- Sortino, Calmar and profit factor display n/m with fewer than three losing months; otherwise displayed values above 10 are capped. An undefined ratio displays a dash.
- Bitcoin benchmark
- Passive Bitcoin buy and hold using month-end UTC prices. Benchmark period returns and drawdowns compound, unlike fixed-capital strategy returns.
- Live versus backtest
- Mean of live and mixed monthly returns compared with the mean of backtest months. Mixed months combine both sources and do not count as full live months.
Twenty-four monthly observations give wide error bands: a Sharpe ratio estimated from two years of data can easily be off by 1 or more, and month-end drawdowns understate the losses open positions can show inside a month. Strategies with many empty months, which only trade on events, have ratios driven by few observations.
Updates
Figures are recalculated once a month after the month closes. The current data runs through Oct 2026.
Not advice
This site shows the operator’s own strategies. It is not investment advice, a recommendation or an offer to manage money.