1. Also, our simulated leveraged proxies from Dec 1969 are somewhat questionable, especially in view of the much higher borrowing rates in the eighties than in the fitted period (mostly from 2009), where most proxy and true (monthly) returns had correlations above 95% between fitted proxies and observed leveraged ETFs. But at least these backtests demonstrates the powerful crash protection of the canary approach (with P2=VWO, BND) combined with asset rotation (as here with the Top5 of the G12 universe), even with very risky (leveraged) assets. 9. Summary and Conclusions We have introduced a variation of our Vigilant Asset Allocation (VAA, see Keller 2017), called DAA: Defensive Asset Allocation. With DAA crash protection is different from VAA by the introduction of a separate protective or "canary" universe. Crash protection is now determined by breadth momentum;one could raise some questions since most of these 2x leveraged ETFs (including UST) are rather illiquid
2. DAA is equivalent to VAA, with its use of asset rotation and selection based on the well-known relative momentum strategy, so EW-Top T (T<=N, long only), without intrinsic or absolute momentum. Compared to this basic relative strength strategy as eg;Otherwise;Faber, 2010 and CXO,2017