One disambiguation before anything else: Quantustik's "entanglement" heatmap is a correlation/dependency visualization borrowed from a physics metaphor — it is not literal quantum entanglement and it is not a prediction that two stocks will move together in the future. With that said, it is genuinely useful for one narrow job: showing you when a portfolio that looks diversified by ticker count is actually one concentrated bet.
Diversification is supposed to reduce risk by spreading capital across positions that do not all move together. The trap is that ticker count is not the same as diversification: five mega-cap semiconductor and AI-infrastructure names can look like five separate decisions on a watchlist while behaving, in practice, like one leveraged bet on the same macro theme — they rise together on the same catalyst and fall together on the same risk-off day.
A correlation view exists to catch exactly this: it tells you, quantitatively, how much two positions' return streams move together, so "I own 5 stocks" and "I own 1 concentrated exposure" stop being indistinguishable.
The live correlation map defaults to the Pearson correlation of each ticker's model forecast path — log-returns of the mean forecast curve — over a chosen horizon. This is model-implied co-movement, explicitly not a claim about realised market correlation. It also ships a "Realised" toggle that swaps in trailing daily-return correlation over recent history — the consensus baseline every correlation tool ships — kept as a separate, clearly labelled mode so the two are never conflated.
Both views degrade honestly when data is thin: a ticker with too few overlapping return points, or a below-cutoff pair, is reported as missing rather than filled in with a guess. Clicking any ticker drills into its single-row view — most and least correlated names against everything else in the current universe.
The practical use is portfolio-shaped, not trade-shaped: pull up your own holdings, check their pairwise correlation, and treat a cluster of high-correlation names as one risk-sizing decision rather than several independent ones. If four of your positions move together, size that cluster the way you would size a single large position — not the way you'd size four small, independent ones.
This is complementary to, not a replacement for, the sector rotation heatmap: two names can sit in different GICS sectors and still be tightly correlated in the model's forecast paths (a supplier and its largest customer, for instance), which is precisely the kind of hidden concentration a sector-only view would miss.
What it is not useful for: picking a pairs trade, predicting the next earnings reaction, or deciding whether a stock will go up. Zero forecasting edge is a perfectly fine starting assumption for this feature — it is a structure tool, not a signal.
The default view correlates model forecast curves, not realised prices — it is model-implied co-movement, not a claim about how the stocks will actually trade. Correlation is also a single number over one recent window; it can and does shift after earnings, M&A, or a sector-wide repricing, so a low-correlation pair today is not guaranteed to stay uncorrelated. Correlations of smooth forecast curves also tend to run more extreme than realised-return correlations — two gently trending curves can correlate near ±1 on shape alone — so read the default view as a relatedness ranking and cross-check the Realised mode before acting. And it says nothing about direction or size — two positions can be highly correlated and both wrong, or uncorrelated and both still lose money in a broad drawdown.
Educational research only — not investment advice.