Jul

5

The calendar here at Daily Speculations puts market days into four groups, based on the daily changes in S&P futures and bond futures:

Green = Stocks Up, Bonds Up
Orange = Up, Down
Blue = Down, Up
Red = Down, Down

Using daily data for the S&P and for TLT, from 2 January, 2024, to 28 June, 2025, I determined which color each day is, and then did the count for each color, and what % that color day is of all days:

Then I counted what follows each day, i.e., a Green day could be followed by another Green day, or an Orange, or Blue, or Red day. With a random distribution of days, you would expect random following days, i.e., if 40% of days are Green, and 30% are Orange, then you would expect any given day to be followed by a Green day ~40% of the time and an Orange day ~30% of the time. You could then look at deviations, e.g., Blue days followed by Orange days only 25% of the time could be counted as -5%-point deviation.

So I did this kind of counting with the calendar days, with these results, where you see, for example, the number of times a Green day follows a Green day (39), what % of the time this represents (33.1%), and the deviation from expected, measured in % points (1.33%).

• What follows Green days looks random (i.e., the numbers in the deviation column are close to zero percent).
• Orange days are somewhat more likely to be followed by Red days and less likely by another Orange day.
• Blue days look random.
• Red days are more likely to be followed by Orange days.

I keep thinking I should study Markov processes, especially "Hidden". I don't know if this kind of counting is a simple version of a Markov process, and if there is more that could be done.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search