Methodology

Walk-forward backtesting

All backtest results use walk-forward validation. At any point in the test, the model uses only data available before that point. This eliminates lookahead bias, where a model inadvertently uses future information to make past decisions.

Results are computed for multiple entry points across the full available price history at weekly intervals. The charts display the distribution of outcomes across these entry points, not a single cherry-picked result.

Data sources

The backtest pipeline uses two primary data sources:

• Price data: Daily OHLCV (open, high, low, close, volume) from CoinGecko, cross-referenced against secondary sources. History extends from 2010 to the present.
• On-chain data: MVRV ratio, SOPR, realised price, supply in profit, and short-term/long-term holder cost basis. Sourced from on-chain analytics providers. History varies by metric, typically from 2012 onwards.

Update frequency

Backtest results are recomputed weekly via an automated pipeline. The web application serves pre-computed results; it does not run backtests in real-time. Each set of results displays a "last updated" timestamp.

Results may lag current market data by up to one week.

Strategies

Four strategies are backtested:

Costs included

All backtest results account for:

• Exchange fees: 0.1% per trade (baseline estimate)
• Bid-ask spread: 0.05% per trade
• Tax impact: calculated per jurisdiction using the rules in the tax-rules package

Actual costs vary by exchange, order size, and market conditions. The estimates used are conservative for major exchanges.

Tax calculations

Tax-aware results apply jurisdiction-specific rules from the tax-rules package, covering Australia, the United States, the United Kingdom, Canada, and Germany. Calculations include holding period discounts (where applicable), capital gains rates, and the tax impact of each taxable event generated by the strategy.

Tax calculations are estimates based on general rules. They do not account for individual income levels, other gains or losses, or jurisdiction-specific exemptions beyond those modelled. Consult a qualified tax professional for accurate tax advice.

Limitations and caveats

• Limited historical sample. Bitcoin has traded since 2010, providing a small number of complete market cycles. Statistical conclusions drawn from this sample size warrant caution.
• Structural market changes. The introduction of ETFs, institutional custody, and regulated derivatives is changing Bitcoin's market structure. Models trained on historical data may not capture these shifts.
• On-chain metric reliability. As more Bitcoin is held through custodians and ETFs, on-chain metrics may become less precise at reflecting true holder behaviour.
• Data freshness. Results may lag current market data by up to one week due to the weekly pipeline schedule.
• Past performance. Historical returns do not guarantee future results. Every strategy presented in this tool could produce negative returns over any given period.