Atlas · Methodology

How every score is built

Atlas scores every stock in the Nifty 500 each night across six lenses, blends four of them into a single 0–100 conviction score, and rolls that same read up to sectors, funds and ETFs. It is a glass box: every number traces back to the inputs it came from.

This page shows exactly what goes into each lens and how the lenses combine. Nothing here is hidden, and nothing is invented — if an input is missing for a stock, that lens is scored on what is present and the rest re-normalises, rather than filling the gap.

The engine

What runs every night

The same pipeline runs every weekday night. Click a step to see what happens.

Fresh inputs land for the whole universe: NSE prices, volume and delivery; quarterly & annual financials plus ready ratios; exchange filings (announcements); insider deals, bulk deals and shareholding; and the latest monthly mutual-fund holdings.
In
NSE · financials · filings · holdings
Out
latest inputs, point-in-time
The score, decomposed

What goes into every lens

Every conviction score is built from six lenses, and each lens from its own sub-components — all computed per stock, every night, and stored so you can trace any number back to its inputs. Two lenses form the blend today; the rest are context. Weights are live from the thresholds panel.

Conviction Composite
0–100, per stock
0.90·Technical + 0.10·Flow
Technical90%
Scored lens
Price action and momentum versus the market.
Trend
EMA 21/50/200 alignment · price vs EMA-200 · RSI-14 · 1-week return
Relative Strength
RS vs Nifty 500 (1m·3m·6m·12m) · RS vs own sector
Volatility Contraction
ATR-14 as % of price · Bollinger-band squeeze
Volume
30d / 60d volume ratio · 52-week price position
Fundamental0% · context
Context · not in blend
Business quality read straight from the financials.
Profitability
ROE · ROCE · net margin
Margins
Operating margin · net margin
Growth
Revenue YoY · EPS YoY
Balance Sheet
Debt / equity · current & quick ratio
Operating Leverage
Growth × margin expansion × low debt
Flow10%
Scored lens
Who is actually buying — the smart-money footprint.
Promoter70%
Insider open-market buys/sells · pledge changes · promoter holding %
Institutional / Smart Money30%
Mutual-fund month-over-month delta · bulk deals · FII/DII shareholding QoQ
Accumulation25%
Delivery % level & trend · up/down-day asymmetry
Catalyst0% · context
Context · not in blend
What just changed — read from exchange filings.
Earnings & Momentum55%
Credit-rating actions · dividends · order wins · press releases
Capital Actions30%
Acquisitions · buybacks · bonus / split
Governance15%
Management & auditor changes · ESOP
Valuation× multiplier
Context · not in blend
Not part of the blend — it tunes the final score up or down.
PE vs Sector
PE relative to the sector median
Absolute PE
Trailing-twelve-month PE
Price-to-Book
P/B
EV / EBITDA
Enterprise value / EBITDA
52-week Position
Where price sits in its 52-week range
Policycontext
Context · not in blend
Government tailwind, sector-matched. Shown for context, not scored.
Government Tailwind
Match vs 15 policy schemes — PLI Electronics/Pharma/Auto, Defense, Semiconductors, Green Hydrogen, FAME-III EV …
Convergence boost
When lenses agree (each ≥ 40), the score is boosted: 2 lenses ×1.06 · 3 lenses ×1.10 · 4 lenses ×1.15.
Valuation multiplier
Deep value ×1.15 · Cheap ×1.08 · Fair ×1.00 · Expensive ×0.90 · Overvalued ×0.75.
Conviction tier
HIGHEST ≥ 70 (3+ lenses) · HIGH ≥ 58 (2+) · MEDIUM ≥ 45 · WATCH ≥ 30 · below that, no call.
Every node, expandable

The whole model, click to open

The conviction score breaks into its lenses, each into sub-components, each into the real metrics behind it — then how a single stock’s read rolls up to a sector, an ETF / fund, and a category ranking. Each metric’s definition is the same one behind its info-icon on the tables. Lens weights are live.

Live values

Weights & thresholds

Read live from atlas_thresholds — the same row the scoring engine reads and the control panel edits. Change a value there and every score, and this page, move together.

Lens weights
Technical0.9090%
Fundamental0.00context
Flow0.1010%
Catalyst0.00context
Valuation & Policy: context (weight 0)
Convergence boost
Applied when scored lenses agree (each ≥ 40):
2 lenses×1.06
3 lenses×1.10
4+ lenses×1.15
Conviction tiers
HIGHEST70, 3+ lenses
HIGH58, 2+
MEDIUM45
WATCH30
One atom, everywhere

How stocks roll up to sectors, funds and ETFs

There is one scoring engine. Everything above a single stock is a weighted roll-up of the same lens scores — no separate model.

Sectors

A sector score is the free-float-weighted average of its constituents’ lens scores. Bigger companies move it more — the same four-lens blend, one level up.

Funds & ETFs

A fund’s lens read is the holdings-weighted average of what it actually owns. Its Technical score is its holdings’ Technical scores weighted by position size — fully traceable to the names inside.

Same scale everywhere

Stock, sector, fund and ETF all sit on the same 0–100 lens scale and the same composite blend, so a number means the same thing wherever you see it.

Limits

What the score is — and isn’t

So you read it for what it is.

A read, not a promise

The score describes what is strong right now across the lenses. It is a transparency view of the evidence, not a forecast that a stock will go up.

Data lags reality

NAVs run a day or three behind, fund and ETF holdings carry SEBI’s ~30-day disclosure lag, and filings are as-reported. Scores reflect the latest available, not real-time.

Missing data degrades, never faked

If a lens input is absent, that lens is scored only on what is present and the composite re-normalises. Atlas never substitutes a fake neutral to hide a gap.

It does not size positions

Atlas ranks and scores. How much to hold, and your own risk limits, are your call.

Every weight and threshold on this page is configurable and lives in the thresholds control panel. The lenses, sub-components and weights shown are read from the live scoring engine and atlas_thresholds.