Subject: Re: Jim, thank you ...
That's actually my point, that I can't see how to assess that from his table starting 2009 as there were not a single one longer than 5 months (apart from the current one).
To assess that one would need a table with the signals going back much further --- but then the question would be whether what was valid 20, 30 years ago still is valid today.

Again, it's important to define "success" in an agreed way.
All of the broad US market returns, and more, occurred in the subset of the time tagged as bullish.
The *average* return during those times was good.
It's certainly true that individual bullish stretches had some unpleasant times, but overall you got all the returns in the intervals expected.
Being in the market the rest of the time was, on average, a low-return proposition. Negative, in this stretch.

On average through the years this signal is bullish 73.2% of the time.
In this post-publication era it has been bullish 83.8% of the time.
For a simple reason: we have seen mostly bull markets since 2008, which it seems to have noticed correctly.

It was originally tuned on data 1950-2008, then validated out of sample 1930-1949, then validated post discovery 2008-2022.

Can it be used as a trading system? Maybe. Long if the market appears bullish, cash otherwise.
If one were in the market only during the bullish times, the data suggest you'd have a portfolio with under 65% of the risk of buy-and-hold and returns about the same or a bit higher.
The risk metric I use ignores volatility. It considers only the chances of having a negative rolling 3 month stretch,
and the chances of having a rolling years stretch with a nominal return under 10%.
There's a squared penalty on bigger shortfalls below those thresholds, and more importance given to rolling years than to rolling quarters.
I don't use the signal this way, and don't advocate it, but it seems to work.
For example, in the data from 1930, the worst rolling two-year period buy-and-hold was a real total return of -35.9%/year compounded.
For the hypothetical long-or-cash person, the worst two years would have been -18.9%/year. Certainly bad, but better.

Jim