The market
is not a list.
We rebuild the NIFTY-50 correlation graph from 1-minute bars and measure where every stock sits inside it. Move the window dial above: between a 30-bar and a 120-bar estimate, 75% of the edges change and the most-central stock is a different name entirely. The nodes that barely move are the ones you can believe.
- Universe
- NIFTY 50
- — tradable names · NSE
- Communities
- —
- Louvain · —
- Modularity
- —
- —
- Most central
- computing…
Not a report. A machine you run.
Every number on this page came out of the same engine you get access to. Set your own window, universe, and metrics — then look.
Graph Stats
Rebuild the correlation graph for every as-of day in your window and track each stock's centrality through it. Five metrics, four graph methods, any symbol set. Export the whole series.
Network Graph
Today's clusters as a force-directed plate. Hover a ticker and its influence neighbourhood lights up hop by hop — a tight cluster flashes at once, a bridge lights in slow, thin chains.
How It Works
The full methodology: from 1-minute bars to correlation graphs, community detection, five centrality metrics, and why we killed the lead-lag product.
Prices in. Structure out.
Most signals stop at the stock. Ours start at the edges between them — and no step is a black box.
Log returns, session-aligned
Every stock's OHLCV is resampled from 1-minute bars to your interval, then converted to log returns over a trailing window you choose. The window is the single most consequential parameter in the whole pipeline, which is why we let you move it and watch what happens.
Correlation becomes structure
Pearson (or Spearman) correlation across the window gives a dense matrix. We sparsify it into a graph — a Mantegna minimum spanning tree by default, or kNN, threshold, or complete. The sparsifier is a choice, and it is yours.
Where each stock sits
Louvain finds the communities; five centrality measures score each stock's position — eigenvector, PageRank, degree strength, betweenness, closeness. Rebuilt per as-of day, with a fixed seed, so the partition does not shuffle between calls.
— tradable names on NSE. TMPV is excluded — it has no price history, and we would rather drop a name than fake one.
Resampled to any interval from 1m to 1d. Nothing is interpolated.
Rebuilt every session, not backfilled. We have no history before 2025 and we do not claim any.
One engine. Two access levels.
The ladder is the engine's own limits — intervals, sparsifiers, window depth, run count. Paid unlocks the wider engine, and a lot more is available on request.
| Parameter | Free | Paid |
|---|---|---|
| Price | ₹0 | TBD |
| Universe | NIFTY-50 | NIFTY-50 · NIFTY-500 · custom † |
| Intervals | 1d | 1m · 5m · 15m · 1h · 1d |
| Graph methods | MST · kNN | MST · kNN · threshold · complete · more † |
| Correlation | Pearson · Spearman | Pearson · Spearman · more † |
| Centrality metrics | All 5 | 5 + more † |
| Max lookback | 100 bars | 300 bars · deeper † |
| As-of days per run | 10 | 10 · more † |
| Symbols per run | 50 | 50 · more † |
| Runs | 5 | Unlimited |
| CSV export | Yes | Yes |
| Network graph | Yes | Yes |
- Wider universes — NIFTY-500 and custom baskets of your own symbols
- More instruments and longer price history
- Additional centrality metrics beyond the core five
- Further graph constructions and sparsifiers
- More correlation estimators (distance, partial, tail-dependence, …)
- Deeper lookback windows and longer rolling series
- Full intraday at scale, and bespoke research runs
Skylife Research provides quantitative research and data, not investment advice. Graph centrality measures market structure — it is a research input, not an entry or exit signal. Trading involves risk.
Skeptical? Good.
The questions a serious desk asks before it trusts anyone's numbers.
What does Skylife Research actually do?
We build and publish a live correlation graph of the NIFTY-50. Every trading session, the engine ingests 1-minute OHLCV bars from NSE, computes pairwise correlations over a trailing window, sparsifies the result into a graph, detects communities, and scores each stock on five centrality metrics. The output is a structural map of the market — who is central, who is peripheral, and how that shifts over time.
What is a correlation graph, and why should I care?
A correlation graph is a network where each stock is a node and each edge represents how closely two stocks have moved together over a defined window. Traditional analysis looks at stocks one at a time; a graph lets you see which stocks behave as a cluster, which ones bridge clusters, and which are genuinely independent. If you hold five stocks that all sit in the same dense cluster, you own one bet with five tickers.
How is correlation calculated?
Pearson correlation on log-returns, computed over a trailing window of your chosen length. We also support Spearman (rank) correlation. The window length is the most consequential parameter in the pipeline — a 30-bar and 120-bar window will share roughly 25% of their strongest edges. We let you move the window and watch what changes, because most vendors pick one and never show you the others.
What does the graph actually look like, and how do I read it?
On the Network Graph page you see a force-directed layout: each stock is a dot coloured by its Louvain community, and edges connect correlated pairs. Hover a ticker and its influence neighbourhood lights up hop by hop — a tight cluster flashes at once, a bridge stock lights in slow, thin chains. The position of a node tells you its structural role; the colour tells you its cluster.
What do the edge weights mean?
Each edge carries the Pearson (or Spearman) correlation coefficient between the two connected stocks over the current window. A weight near 1.0 means the two stocks moved almost identically in that period. After sparsification (MST, kNN, threshold, or complete), only structurally significant edges survive.
What are the key metrics you compute, and what do they tell me?
Five centrality metrics per stock per as-of day: eigenvector centrality (influence through influential connections), betweenness centrality (how often a stock sits on shortest paths between others), PageRank (recursive importance), degree strength (total weight of all edges), and closeness centrality (how quickly information could spread from this node). Together they answer: is this stock a hub, a bridge, a satellite, or an outlier?
What is eigenvector centrality, and why is it your “flagship” metric?
Eigenvector centrality scores a stock based not just on how many connections it has, but on how important those connections are. A stock connected to five central banks will score higher than one connected to five peripheral chemicals. It captures systemic influence — the stocks that, if they move, tend to pull the whole market with them. It’s our flagship because it best reflects structural importance.
What is betweenness centrality, and how is it different?
Betweenness counts how often a stock lies on the shortest path between every other pair in the graph. High betweenness means the stock is a bridge — it connects clusters that would otherwise be separate. These bridge stocks are structurally important for a different reason: they transmit shocks between sectors. A stock can have low eigenvector centrality (not in the central cluster) but high betweenness (it connects the central cluster to a peripheral one).
How should I use the daily pulse on the homepage?
The pulse is a live reading from the engine, not a signal. It tells you the current state of the graph — how many communities the market has resolved into, the modularity score, the most-central stock, and the most-drifted stock. Use it as a structural check: if the market suddenly drops from 5 communities to 2, correlations are spiking and diversification is collapsing. That’s information.
Can this predict the market?
No — and we would rather say so. Graph Stats measures market structure: who is central, who is peripheral, and how that shifts. It is a research input, not a prediction or signal. If someone sells you centrality as alpha, they are skipping several steps. Our job is to show you the structure honestly, including when it is unstable.
Where does the data come from?
1-minute OHLCV bars from the National Stock Exchange (NSE) of India, sourced via Angel Broking’s SmartAPI. The data is for the NIFTY-50 universe — 49 tradable names (TMPV, which has no trading history, is excluded). Data starts from January 2025. There is no history before that, and we do not claim any.
How is the correlation window chosen?
You choose it. The window is the number of trailing bars used to compute each correlation matrix. A short window (e.g. 30 bars) captures recent regime shifts but is noisy. A long window (e.g. 120 bars) is stabler but smooths out structural changes. There is no universally “correct” window — which is exactly why we let you set it and compare.
What is the Louvain algorithm, and is the clustering stable?
Louvain is a community-detection algorithm that maximises modularity — it finds groups of stocks that are more densely connected to each other than to the rest of the graph. It is fast enough to rebuild the graph per as-of day, which makes the time-series product possible. We run it with a fixed seed, so the same window always yields the same partition. Modularity (Q) is reported with every result so you can judge how real the separation is.
Do you do lead-lag testing — who moves before whom?
We tested it. Granger causality across all 2,450 directed pairs in the NIFTY-50, with multiple lag structures. After Benjamini-Hochberg false-discovery-rate correction at 10%, zero pairs survive. So there is no lead-lag product, and nothing on this site shows a directed flow between two stocks, because that would imply a relationship our own research rejects.
What is Granger causality, and why doesn’t it work here?
Granger causality tests whether past values of stock A help predict future values of stock B, beyond what B’s own past predicts. In theory it could reveal lead-lag relationships. In practice, with 2,450 pairs and multiple hypothesis tests, any apparent leads are indistinguishable from noise after FDR correction. We could have shipped it anyway — most vendors do — but it would be selling a false positive.
What is the false discovery rate correction, and why does it matter?
When you run 2,450 statistical tests (one for each directed pair), some will appear “significant” by chance alone. The Benjamini-Hochberg procedure controls how many of your claimed discoveries are actually false. At a 10% FDR threshold, you accept that up to 10% of your reported pairs may be spurious. Even with this lenient threshold, no pair survives — which means the lead-lag signal in the NIFTY-50 is indistinguishable from noise.
What is Portfolio Overlap, and how does graph-based overlap differ from traditional overlap?
Traditional portfolio overlap counts shared tickers between two portfolios. Graph-based overlap goes further: if two stocks are in the same Louvain community with high mutual correlation, your portfolio carries concentrated risk even if the tickers are different. Upload your holdings and see how much of your book is the same structural bet wearing different names.
Can graph centrality replace traditional diversification analysis?
No — it supplements it. Traditional diversification looks at sector allocation, correlation matrices, and VaR. Graph centrality adds a structural layer: it shows you when stocks that look different (different sectors, different betas) actually occupy the same node neighbourhood in the correlation graph. The two approaches answer different questions, and both are worth asking.
How is Skylife Research different from a stock tip or signal service?
We do not tip stocks, recommend trades, or provide entry/exit signals. We publish quantitative structural research: the correlation graph, centrality metrics, and community detection results for the NIFTY-50, rebuilt every session. What you do with that information is your decision. The disclaimer is not a legal footnote — it is the product philosophy.
What markets or exchanges does Skylife Research cover?
Currently, the NIFTY-50 on the National Stock Exchange (NSE) of India. On request, we run wider universes (NIFTY-500, custom baskets), longer history, additional metrics, and alternative graph methods. These are gated behind a conversation, not a self-serve toggle — tell us what you need.
Ask us anything.
Questions about the methodology, the data, API access, or the paid tier — or a wider universe and features on request. We read and reply to every message.
If you want to evaluate this for a desk, bring one portfolio and a window length you trust. We will run the graph on it live and show you which of your positions are the same bet.
- Based
- Mumbai, Maharashtra, India