The evidence behind the models
What changed the model, what failed the test, and what remains an assumption.
These 13 studies cover strategic allocation, KISS signals, taxable portfolios and diversification. Each article identifies its original study period and decision. Publishing an older study does not turn its backtest into a live track record.
October 8, 2026
A reconstructed equity and liquidity proxy explains some bitcoin risk, but unstable forecast results leave its expected return unresolved.
Exploratory; no forecast adopted
Read study →October 8, 2026
A fixed approximation of ELM's underlying allocations shows how internal cash, forecast uncertainty and bitcoin exposure affect portfolio risk.
Conditional allocation and risk diagnostic
Read study →October 6, 2026
Why CAGR needs converting, why 200-week-average forecasts were not adopted, and how Central and Robust allocations respond to uncertain inputs.
Implemented, with rejected forecast models
Read study →October 6, 2026
Correcting the rolling short-TIPS return estimate and separating nominal portfolio optimization from purchasing-power protection.
Implemented return correction
Read study →September 29, 2026
A lower-volatility mix of stocks, nominal bonds, TIPS, gold and commodities, tested with common dates, costs and execution assumptions.
Retrospective comparison
Read study →September 29, 2026
Replacing part of IEF with SCHP modestly improved these historical portfolios. The study also shows what the comparison cannot isolate.
Diagnostic study
Read study →September 8, 2026
Eleven rebalancing policies, explicit tax lots and a liquidation check show how taxes, missed recoveries and drifting bitcoin exposure interact.
Research only; no new rule selected
Read study →September 2026
A monthly trend and drawdown rule helped in prolonged equity bears, but failed as crash insurance and as a shared bitcoin exit rule.
Historical candidate study
Read study →September 7, 2026
Correcting basket weighting, neutral evidence and missing data reduced the backtest return. Those fixes still belonged in the model.
Construction fixes adopted; candidates held back
Read study →September 2026
A controlled synthetic test and a separate historical gate comparison reveal different costs of smoothing and removing trend signals.
Replacement rejected
Read study →July–September 2026
Two closes to reduce risk and five to add it cut trading in the original study. Later tests exposed counter interactions and the cost of faster entry.
Confirmation policy adopted
Read study →July–September 2026
Gold ETF handoffs, a credit proxy and breadth limitations determine what the historical KISS record can reasonably claim.
Data methodology adopted
Read study →2018–2025 sample; original KISS 2.0 study
The original reconstruction traded some upside for smaller drawdowns. Its frozen assumptions explain why newer comparisons differ.
Historical baseline
Read study →