I prefer to think of it this way:

I didn't call the bottom.
I created a model which tries to do so.
Since then, I merely report what it says, and how it has done.
It's a subtle difference...


FWIW, a long time ago I also published a very simple market timing system that attempts to identify ongoing bull markets.
It tries to identify stretches that are relatively safe to go in the water, leaving other stretches as "not so sure".
Again, a subtle difference between "not so sure" and "the model thinks the market is going to fall" bearishness.
The most useful thing about the signal is that it doesn't change states very often.

For whatever it's worth, these are the results for SPY since the model was posted 14.3 years ago.

The model has been bullish 83.8% of the time.
CAGR for SPY during those days +12.92%/year rate, total return 326.3% not annualized.

"Not so sure", 16.2% of the time.
CAGR for SPY during those days -1.01%/year rate, total return -2.3% not annualized.

It has had slightly less than one signal per year.
(i.e., on average one bullish stretch plus one not-so-sure stretch about each two years)

It hasn't been bullish since 2022-05-26

If you are more of a Nasdaq person, the returns while this model was not bullish were very similar on QQQE (Nasdaq 100 equal weight) at -1%/year.
But the CAGR during the bullish stretches was 3%/year higher at a rate of 15.9%/year.

Bottom line: it's impossible to build a reliable market timing model.
However, it's not impossible to build a model that has statistically significant predictive power.

The trick is to find a way to use that information so that you get an edge from it on average over time,
while not taking on increased risk when it is sometimes inevitably wrong.
You can do very well with even slightly loaded dice, provided you don't bet the rent on every roll.

Jim