BTW, the DBE is triggered 99 days after the most recent 252 day high.

FWIW, the original no-new-high-lately signal was 99 days since trailing 99 day high, using intraday highs, and that signal was modified/adapted when the BCC triple was created, which is fine.

The 99 day approach seemed quite resilient even in turbulent markets like the 1930s, which was used to tune it, since the original goal was "pretty darned sure we're still in a bull", vs "not sure". In recent decades, a timeout of 110-120 days might have been better as a timing signal, as cycles have been much smoother. Other than the Covid plunge, of course.*

As an aside, the old Sjuggerud 1-2-3 model has been not so bad. It's interesting because it looks at three very different things that can affect market direction: momentum, monetary environment, and valuation. Each sub-signal is bullish around 2/3 of the time. It probably wouldn't be too hard to come up with improved substitutes for the individual models, perhaps CAPE instead of P/E etc. But given the long run up-trend of market valuations, you'd probably need something like "CAPE above 70th percentile in last N years" rather than CAPE>constant.

Jim

* Perhaps all momentum models should have a health check: do not sell on a negative crossover if the market dropped more than XX% in the last X days: based on 1987/2009/2020, true crashes are often best seen as buying opportunities (if you spot the bottom) or "just keep holding" opportunities if you didn't. If you reduce the pain of one gigantic whipsaw every decade or two, most timing models would give a much better long run result.