I decided to play around with some data today...
Listened to a podcast where Cathie Wood was talking about companies that innovated and how they were fighting for a share of the pie ("winner takes most"). It made me think about the interdependencies of the market. Past performance is used to predict future performance, but I am not sure how often models account for how the past performance of "Company B" impacts the future performance of "Company A".
For my job, I regularly use a type of network analysis that is essentially based on graph theory. We have nodes ("common stocks") and edges ("magnitude of association between stocks"). How does "Company B's" changing performance impact "Company A's" performance? We can conduct what is called a fixed effects analysis, which examines within-company change in stock price and its impact on another companies change in stock price. More importantly, we can lag this relationship. How does the performance of "Company A" on a Monday influence the performance of "Company B" on a Tuesday (for example)?
I took a look at the ARKQ ETF, which focuses on innovation in robotics/autonomous vehicles. I created an Excel file that gathered the closing price data over the last 180 days (125 days in which the market was open, so 125 data points) for all 42 common stocks in this portfolio. For ease of visualization, what I'll show here are just analyses of the 10 highest-weighted holdings and the 20 highest-weighted holdings in the ARKQ portfolio.
When you see an autoregressive edge (an arrow going back into the node), it indicates that past positive performance predicts future positive performance. Whoop-dee-do.
What we are interested in are the edges that show a directional arrow between two nodes (stocks). It will be a bit hard to see all arrows, but this is by design. The arrows that are more heavily weighted represent a stronger effect size. The heavier the arrow, the stronger the impact of a change in price for "Company A" on the change in price for "Company B" during the next day of trading.
Arrows that are blue indicate a positive fixed effect. What this means is that the effect of a change in "Company A" results in a change in the same direction for "Company B" during the next day of trading. If "Company A" is up on Monday, "Company B" will be up on Tuesday. Arrows that are red indicate a negative fixed effect. The effect of a change in "Company A" results in a change in the opposite direction for "Company B" during the next day of trading. If "Company A" is up on Monday, "Company B" will be down on Tuesday.
I am sure that people will know that I am just messing around/having fun with data, but to be as clear as possible, obviously do not take any of what I am doing as advice/recommendation/endorsement.
In theory, FLIR was down today. If the model is accurate, DE will be up tomorrow. Similarly, TWOU was down today, so KMTUY and CAT should be up tomorrow (see the model with the 20 nodes).
The nodes are not randomly placed in the graph. An algorithm is used to push towards the center those nodes that are more central; i.e., more information flows through them. DE, KTOS, and TSM are especially informative of how other stocks will perform (typically, if they are up, the next day, a lot of other stocks will be up).
Anyway, that was just for my own entertainment and I wanted to spew my thoughts out, so apologies for spamming the thread with my "Dear Diary" moment.