Subject: An analysis of my timing signals
With some of my newly found time from a forced early retirement, I have done some shallow statistical assessments of the predictive value of the timing signals I track on a weekly basis. With Google Gemini’s assistance.

I have tracked about 23 timing signal values weekly since March 2019. The signals are a mix including the (four) historical BearCatchers, several technical chart signals such as DMI, PPO, and some breadth/warning indicators, including a few of Zeelotes’ published items, using cheap paid or free data sources. The signals are intermediate term or short term.

Here are some assessments.
There is nearly zero statistical validity regression-wise in predicting the value of the S&P forward 1 or 2 month returns, in either of two metrics:
* The net total of Bullish - Bearish signals (which has oscillated between -15 most negatve and 15 most positive)
* The 1 week or 2 week change in that net total.
This should be kind of expected as we have been in a bull market for most of the duration of time since early 2019 - 70-73% of the 1 month and 2 month returns have been positive.

However, there are a few useful insights.
The highest median and average 1 month and 2 month (especially) forward returns were when the signal state was bearish to very bearish. (in the bottom 30% of dashboard values ie below -6).
Gemini: “Overall, both "1 month chg" and "FWD 2 mth chg" tend to be higher in the lowest decile (D1) of "TOTAL" values and lowest in the mid-range deciles (D6 and D7). This suggests that very low "TOTAL" values might be associated with slightly higher future changes, while middle-range "TOTAL" values might be associated with lower future changes.”

So, A neutral dashboard has indicated neutral forward short term returns can be expected.
Also, this seems to reflect the pretty quick snapbacks / v-bottoms we’ve seen from extreme lows.

Similarly, the highest percentage of positive forward 1-2 month returns (88%, 70% & 75%) was when the dashboard is bearish to very bearish (bottom 2 deciles) or very bullish (top 2 deciles) - at either end of the spectrum.

On the efficacy of using NH/NL BY ITSELF (in THIS timeframe) as a bearish indicator worthy of selling,
77% of the times that the NH/NL flipped to bearish in a given week, the S&P was up from that point both 1 and 2 months later.

Drawdowns in the last 6.5 years

I looked at the instances where the S&P was in a drawdown of more than 8%. Obviously, the 20 covid bear and the 22 interest rate bear make up most of those occurrences. Then I looked at the state of the intermediate signals 4 weeks before those drawdowns.

The most consistently bearish signals in advance of that significant drawdown were
1.) the S&P PAMA 200 (breadth) below 60% (bearish 93% of the time 4 weeks before that level of drawdown),
2.) the SMA Slope bearcatcher (bearish 81%),
3.) the market being at least 7% off its high (78%) - indicating that “beyond correction” level had a very high chance of getting worse
4.) Tie: Nasdaq bullish % breadth indicator bearish (74%), and NH/NL indicator was bearish

Those 4 together could be a decent “no, really, Get Out” indicator. Then again, there's some noise in this data (the S&P's been dominated by the big 7-10 for a long time) so who knows. I just try to play defense when things look red and preserve some capital 30 years into serious investing.

Plug - I’ve gotta say that Google’s Gemini AI tool is amazing at its ability to do analysis on Sheets. The drudgery of figuring out spreadsheet formulas, selects, selective table copies or Python scripting is gone (because it does all that as your prompts require). It provides user friendly stats 101 explanations. For the great unwashed, it’s a fit. For data scientists & mathematicians, well there’s R and Python (which Gemini uses under the covers.)

Regards,
FC