Posted this video on X/Twitter under @vecvizanalytics pre-market on the morning of 5/15/25 with the comment “$UNH chart looks like TDG circa March 2020 as per Vector Model chart shape metrics. That would be a good thing bulls over 3m-6m. Conversely, if UNH has truly “leveled down”, we show the case for settling in at the $225-$260 zone. All in the video below.”
Note that post close on the 14th the WSJ article about the DOJ investigation came out, and that pre-market on the 15th the stock was near $280. The $225-$260 we cited in our comment was based on narrative linked channel analysis. The $225- $260 range was well below the pre-market price and the Vector Model’s 99D (99%tile VaR) level for the 3 month and 6 month forward dates in the dashboard analytics table. UNH got into the mid $250’s later on the 15th, but rebounded and sits at $295 at present.
We intend to post case studies on X/Twitter first, to establish a “time-stamp”, then add the content here.
Enlarged snippet of price probability percentile table featured in the video:
What are all those grey lines and how does all this relate to established techniques and concepts of quantitative finance?
Those grey lines are the core of the Vector Strength Histogram. Each individual grey line is part of a Vector Set. Think of a Vector Set as a triple decker Fibonacci price channel comprised of 15 lines: one touching anchoring top(s), one touching anchoring bottoms, and three in the center. Then we add a channel of similar width, also featuring three center lines, above and below. All lines in a Vector Set are extrapolated forward to the Model Date and linked to a horizontal bar thereafter that represents their Vector Strength.
Vector Strength is our scoring of the “support and resistance” each Vector Set represents to the model date price. Our criteria of touch count and proximity in time and price terms correspond to the findings of academic research.
A full Vector Strength Histogram, like the one featured in the video and also in the right image, below, can be made up of a lot of these Vector Sets – up to 325 of them, in fact. The correspondence of the Vector Model’s price probability percentiles (blue) for 6 forward dates, up to one year out into the future, to the Vector Strength Histogram is often clear when it is overlayed upon it. Sigma’s price percentiles are overlayed as well, for contrast.
Cumulative Vector Strength traversed is how the Vector Model scales price movement. In other words, it is how the Vector Model scales price volatility. The more Vector Strength between the current price and some other price, the further away it is in terms of likelihood of reaching it, all else (especially chart shape metrics) equal.
There are several parallels between Vector Sets and our use of “Vector Strength” to scale volatility, and techniques and assumptions commonly employed in traditional quantitative finance:
1) Emphasis on More Recent Data: Ascribing greater Vector Strength to recently formed Vector Sets is akin to the well-established principle in quantitative finance that “sigma” (standard deviation) based volatility metrics are improved by applying exponential decay to the lookback window of returns. This gives more weight to recent data.
2) Scaling of Deviations in Comparable Quantities: Vector Strength is assigned to Vector Sets, which are defined by at least one price top and at least one price bottom. Consider a simple case: a flat Vector Set anchored by one top and one bottom. Assume the distance between this top and bottom—the “core channel” of the Vector Set—represents one “standard” deviation (conceptually, rather than a strict statistical definition) both above and below the channel’s center. By constructing each Vector Set with a “core channel” and approximately equally sized “upper” and “lower” areas (or “levels”), each Vector Set aims to encompass a range equivalent to roughly six standard deviations (e.g., +/- 3 sigma from the mean). This range essentially captures the full spectrum of variability often considered in models like Black-Scholes (at least for a one-year forward period from the dates of the anchoring top and bottom).
3) Application of Central Limit Theorem Principles: The Vector Strength of a single, isolated Vector Set may have limited meaning. However, VecViz generates hundreds of Vector Sets for each ticker it analyzes. Each Vector Set can be considered an estimate of the range of price deviation, based on a sampled period of price history (defined by the dates of its tops and bottoms). By aggregating these Vector Sets on a Vector Strength-weighted basis, VecViz attempts to estimate the true forward one-year range of deviation for the ticker. The underlying intuition is that a sufficiently large collection of samples can reveal characteristics of the entire population, which is a core concept of the Central Limit Theorem.
4) Parallels with Factor-Based Risk Models: The number of times a Vector Set’s boundaries (lines connecting tops or bottoms) have been touched by price is an influential component in its Vector Strength calculation. A Vector Set boundary line with many touches may function similarly to a regression line for the period spanned by its anchoring tops and bottoms. Factor-based risk models in quantitative finance are fundamentally centered on identifying such relationships. However, instead of using lines connecting price tops or bottoms (as VecViz does in constructing Vector Sets), these models typically use time-series data from factor indices (or, less commonly, principal components). Such models map a ticker’s historical variability to its exposures to these factor indices, essentially based on the goodness-of-fit of a regression between the ticker’s returns and the factor returns. They then use the most recent volatility measures (typically sigma) of the identified factor indices, along with their correlations, to estimate the ticker’s forward variability.
Finally, there is also academic research related to the use of neural networks in conjunction with chart imaging that supports several of the “chart shape” features of our Vector Model and V-Score methodology.
But how well does this all work?
You can review the performance of the metrics included in the video above for our 3+ year out of sample performance period in the Reports section of our site.