A simpler trend signal changed direction less often in the controlled test, but it also allowed a larger loss. That was enough to reject it as a replacement for VAMS under the study’s stated criteria.
Two different pieces of research inform this question. One used a deterministic synthetic market fixture. The other removed signal gates from a historical KISS reconstruction. They answer different questions and should be read separately.
A controlled synthetic replacement test
The VAMS validation report compared the existing system with a simpler combination of absolute trend, multiple return horizons and realized volatility. The fixture contained several deliberately constructed regimes and 49 sampled signal observations.
| Synthetic fixture | Signal changes | Maximum drawdown | Ending value, starting at 1 |
|---|---|---|---|
| Existing VAMS | 15 | −25.7% | 1.080 |
| Simpler candidate | 8 | −33.7% | 0.860 |
The replacement reduced changes but failed the requirement that drawdown should not worsen. These are simulated fixture outcomes, not an ETF backtest, CAGR estimate or live performance record.
The review also found that the five trend votes overlap. Their pairwise correlations ranged from roughly 0.25 to 0.88, so five agreeing votes should not be interpreted as five independent pieces of evidence. Parameter-neighborhood checks showed high agreement, but that alone did not establish better portfolio behavior. RSI remained display-only.
Removing gates in a historical portfolio
The September signal review separately reconstructed January 2020–September 4, 2026. It compared the full KISS 2.3 policy with one gate removed at a time:
| Historical reconstruction | CAGR | Maximum drawdown | Annual volatility | Events |
|---|---|---|---|---|
| KISS 2.3 control | 15.11% | −8.07% | 9.45% | 144 |
| Remove macro caps; retain sleeve VAMS | 15.02% | −12.07% | 10.77% | 109 |
| Remove sleeve VAMS; retain macro caps | 20.64% | −14.12% | 13.11% | 52 |
Removing the sleeve gate captured more upside and exposed the portfolio to larger losses. Removing macro caps produced little return benefit and a deeper drawdown. Macro inputs themselves include VAMS-derived evidence, so the final row is not a complete removal of every VAMS calculation.
The table is a retrospective diagnostic using the then-current data vintage and the engine’s prior-close execution-price proxy. It does not establish that one gate independently caused a particular amount of protection in live trading.
The later signal-construction review adopted calculation fixes and held back new candidates. Lower turnover is useful, but a replacement must also preserve the risk-control behavior it is meant to provide.