Waiting longer to buy back risk reduced trading in the original KISS study. It did not improve returns in every period or portfolio variant.
The adopted KISS 2.1 policy required two confirmed closes to reduce risk and five to add risk, with execution on the next trading session. KISS 2.3 retained those thresholds while tracking each asset sleeve independently.
The original comparison
The confirmation validation reconstructed January 2018 through December 2025 using adjusted ETF returns, Treasury-bill cash and 5 basis points each of trading cost and slippage.
| 60/30/10 portfolio | Allocation events | CAGR | Maximum drawdown |
|---|---|---|---|
| Two closes out / two in | 262 | 11.60% | −10.39% |
| Two closes out / five in | 149 | 12.54% | −9.22% |
Events fell 43%. Four-, five- and six-close entry policies produced broadly similar behavior. Six closes had the highest full-period return, but the study retained five rather than selecting the historical peak.
The counterexamples matter. In 2023–2025, five-close entry reduced CAGR from 20.54% to 19.10%. In the 40% stock / 30% gold / 30% bitcoin variant, full-period CAGR fell from 16.48% to 16.15%, while maximum drawdown improved from −21.52% to −19.52%. This was a turnover policy with an opportunity cost.
Why each sleeve needs its own counter
A shared portfolio counter could reset a pending bitcoin increase whenever gold’s target changed, even if bitcoin’s target remained unchanged. The KISS 2.3 correction gave each sleeve its own confirmation progress. Cash absorbs the temporary difference when one side of a rotation confirms first.
The September responsiveness audit also found that confirmation at multiple stages can make the total wait longer than the allocation rule alone suggests. In controlled fixtures, stable entry took six closes and exit three. A strengthening target that repeatedly reset confirmation took longer still. Those are diagnostic cases, not a guarantee of live execution timing.
Why the faster candidate was rejected
A later test tried replacing overlapping confirmation with a faster two-close policy. On the corrected signal construction, the September 7 experiment produced:
| January 2020–September 4, 2026 | CAGR | Maximum drawdown | Events |
|---|---|---|---|
| Existing confirmation | 14.69% | −8.82% | 144 |
| Faster candidate | 14.42% | −11.07% | 238 |
The faster rule spent less time in cash but traded more and suffered a deeper drawdown. It was not promoted.
All these historical tests reuse inspected data and current-vintage macro inputs. They support keeping a simple, explicit waiting rule; they do not establish an optimal universal number of confirmation days.