Data
- Public backtests use dated score rows already present in the connected workbooks.
- Blank rows, future placeholders, and formula errors are excluded rather than coerced to zero.
- Source tabs are disclosed alongside each series.
Signal
- Series include Medium-Term Trend, Long-Term Trend, MRPI, and the Bitcoin valuation composite.
- Thresholds are adjustable for inspection, but the dated scores themselves are not altered.
Execution Assumptions
- The public app does not convert score changes into assumed trades, fills, or P&L.
- Closing-data timing, slippage, fees, and transaction costs remain undisclosed unless a benchmark test is added.
- No backtest on this page should be read as realized performance.
Test Design
- Observed history is limited to the dated rows published in the forward-testing sheets.
- In-sample, out-of-sample, and walk-forward procedures are not fully disclosed in the current repo.
Revision Control
Material parameter, threshold, and methodology changes are logged before a
revised public backtest is published.
Robustness
- The current public page supports threshold and window inspection.
- Alternative lookbacks, sensitivity analysis, and subperiod studies are not yet published.
Limitations
- Hypothetical analysis is hypothetical.
- Look-ahead bias, overfitting, data-mining, and regime dependence must be discussed explicitly.
- Past performance does not guarantee future results.