Aloha Everyone, I'm starting this thread to discuss ideas about nested fractals, inspired by this post by @Sprout Why would it be important to talk about multiple timeframes converging? If you where looking at a line chart, the closing price would be all you see. So, if you're basing your fractal based trading decisions on closing prices, there are certain times when there are more people watching that others, and making decisions at. The table below is a zoomed out version plotting fractal length/period on the vertical (lengths 1 min to 60 minutes) and time of day on the horizontal. Blue marks indicate the alignment of fractal length and time of day. Red plot at the bottom is a heat map of # of timeframes aligning at that time of day. Yes, I know that it's unlikely that anyone is following something random like a 17-minute timeframe, but I wanted to share this plot to show the distribution across 1 to 60 minute timeframes. The table below is the same setup, but only considering what I believe are the most popular timeframes (left axis). Note the two heat maps at the bottom. The upper one being for all occurrences, the lower one being 'key' nodes occurrences. I think of these key nodes as somewhat of a distribution of modes (not means). These key nodes are marked to visually filter out nodes with little convergence. The lower left corner of the table below has a number representing a cutoff for what I think of as a minimum number of nodes converging to be important. From the heat map of key nodes above, the first thing that jumps out to me is the distribution of the number of converging nodes, with the middle of the hour +/- 10 minutes having more eyes on it than on the hour. Although all hours of the day are slightly different regarding number of occurrences, they primarily follow this pattern. As you would expect, there are the most eyes on the hour, followed by the half hour (according to the timeframes I think are most watched). Visually, I see the distribution of the 'key' nodes (two images above) as the opposite of the distribution of all of the nodes (above). In other words, the weighting seems to be around the middle (the half hour mark) of the 'key' nodes vs. around the sides (15 and 45 minute marks of) all nodes. Why did I star this thread? I would like to get others views about fractal nesting. Not just what I'm talking about here, but about anything fractal nesting you think of. Finally, if you do not believe in the validity of fractals, that is OK and I respect your opinion. I just ask that you do not voice that opinion in this thread. Mahalo!
took me a long time to understand what you really are talking about. I will give a test and see, I haven't had any experience with fractals
Great post, your plots are interesting. Curious if you have the ability to map via polar coordinates. oh matplotlib, to know your ways..
I will note, however, that this thread doesn't have to be about nesting price. I believe nesting time is equally (if not more) important.
Now that you've laid out the theory, can you show us how this would apply in practice? BTW you're not merely suggesting we stack 2 timeframes (or however many you like) on the same chart like this, are you? 5m/30m Overlay
Aloha schizo! Not quite. I’m thinking more along the lines of set waves. Theory laid out, but yet to be demonstrated. Next I’m looking for an S/R relationship around these ‘key’ closing prices as well as trying to tie it into volume and price surges. Plenty of work to be done, but inspired by sprouts post, I wanted to start the ball rolling and get others opinions on how they view what I’m thinking.
I have this PhD thesis bookmarked for multiscale financial time series analysis. https://digital.lib.washington.edu/researchworks/handle/1773/48806 It is heavy stuff though and from 2022 so quite new. The other multiscale area of research seems to be in human biological time series but I am over my head on the biology side almost instantly. Multiscale entropy (MSE) is the tool that seems most interesting as there are good implementations of it in R and Python so you just have to understand the assumptions to use it. I haven't been able to grok any of this though. Market profile "other time frame" participants always seemed like a rather obvious market property. How to apply this strategically, I have no idea. With calling these nodes I am trying to picture if this could be represented by a graph, specifically a tree, but I can't picture what the edges would be.
Aloha lariatier, Thank you for sharing. That's quite the read (i very quickly skimmed it). Complex to say the least. I'm hoping to find something simpler. We'll see... lol I'll follow up later with a chart showing one possible explanation of why lunch hour trading has such little participation (other than people needing to eat). Spoiler alert, not a lot of timeframe alignment