How NitiMarg scores countries: methodology

1. Candidates

2. Pillars & weights

3. Normalisation & scoring

For each indicator, take the most recent non-empty year per country from the World Bank (last ~12 years). IMF values use the current year and the next four.
Size variables (GDP, GDP added, high-tech exports, military spending, population, GDP per capita) are log10-transformed, because they span orders of magnitude.
Every indicator is min-max scaled to 0–100 across all candidates (66 since the October 2026 top-50 expansion). Ratio and growth variables are winsorised at the 5th/95th percentile so one outlier can't flatten everyone else. The 65+ share is inverted, so younger scores higher.
Pillar score = weighted mean of its indicators. If an indicator is missing, its weight is spread across the others in that pillar.
Future Power Score = Σ pillar score × pillar weight (0–100). Scores are relative to this candidate set, not absolute.

4. Live GDP estimate

Latest World Bank nominal GDP (US$) × (1 + IMF WEO real GDP growth) for each year up to the current year = full-year estimate. The ticker shows that figure × the share of the calendar year (UTC) elapsed. It is an illustrative estimate, not official data. Real growth leaves out inflation and exchange-rate moves, and output isn't spread evenly through the year. The IMF's own nominal US$ forecast is shown next to it for comparison.

5. Country Exchange: the index-value model (points)

Illustrative model · not a security · not financial advice

Each top-50 country is shown like a listed company. Its "index value" (shown in points, not rupees or prices) is a transparent model, re-anchored to fundamentals on every refresh. No trading, opinion or hidden inputs are involved.

base = Power Score × 10 market_pts = 0.5 × clamp(index daily change %, −10, +10) attention = clamp(articles_24h ÷ median articles_24h of the top 50, 0.5, 2.0) news_pts = 0.2 × clamp(GDELT avg tone 24h, −5, +5) × attention change % = market_pts + news_pts points = base × (1 + change % ÷ 100) index weight = points ÷ Σ points of all 50 × 100

Index daily change = last delayed index level vs the previous session's close (Yahoo Finance chart API; Moscow Exchange ISS for Russia). On weekends and holidays this is the last session's move.
GDELT tone = article-weighted mean of the GDELT DOC 2.0 tonechart for the country query over 24 hours (roughly −10 very negative to +10 very positive). Volume = the number of articles in that chart. Tone older than 36 h is ignored, so news_pts = 0.
If an input is missing, its component is 0 and the page shows "n/a". Nothing is imputed.
Every refresh appends [time, price, change %] to data/history.json. Sparklines are drawn from that stored history (max 2,000 points per country).
Typical daily moves are small (±0–3%) by design. The index value mostly reflects long-run fundamentals (the Power Score), and markets and news only nudge it.

6. Live data and refresh

Feed	Source (free)	Refresh
News + tone/volume	GDELT DOC 2.0 API (English, last 24h, deduplicated by URL and title). If GDELT is unavailable: Google News RSS, clearly labelled as a fallback.	every 15 min (cron)
Trending searches	Google Trends daily trending RSS (top 10). Shown as "unavailable" where Google has no feed (e.g., China).	every 15 min
Stock indices	Yahoo Finance chart endpoint; Moscow Exchange ISS (Russia). Delayed, with the as-of time shown.	every 15 min
Weekly top news (Guide tab)	Google News RSS, last 7 days, India edition (en-IN). Headlines are shown exactly as published, not rewritten.	each country about every 12 h (25 countries per run)
Exchange rates vs rupee	ECB reference rates via frankfurter.app; other currencies from open.er-api.com (Rates By Exchange Rate API). Daily reference rates, not live dealing rates.	daily
GDP, growth, inflation (Guide tab)	World Bank API (latest actual year) and IMF WEO DataMapper (current-year forecast, labelled)	weekly (sources update a few times a year)
Index values + history	Computed by `scripts/refresh_live.py`	every 15 min
Browser	Pages re-fetch `data/*.json`	every 60 s

India lens. Each headline gets a line on how it relates to India. If an OpenAI-compatible API is configured (env OPENAI_API_KEY, optional OPENAI_BASE_URL/OPENAI_MODEL), the top 3 stories per country get a one-sentence note written only from the headline, marked llm:model. Otherwise a keyword rule tags the headline (Mentions India, Trade, Defence, Neighbour) and marks it auto-tag. Tags describe the headline. They are not analysis.

Business niches (top 10 countries) are hand-written ideas grounded in cited World Bank, IMF, MEA, PIB and Reuters sources. They are not advice. The rest are marked "coming soon".

6a. From top 30 to top 50 (October 2026)

Same formula, wider field. The candidate pool grew from 40 to 66 countries: every other economy with nominal GDP above about US$100 bn (IMF WEO), plus large regional or military powers and India's key neighbours. The top 50 by Future Power Score are shown. No country was added or removed by hand.
Because every indicator is scaled across all candidates, adding countries shifts every score slightly. India stays #3.
Just outside the top 50: Kuwait, Portugal, Pakistan, Algeria, Greece, New Zealand, South Africa, Iran and others (see section 1).
Taiwan and Hong Kong are not scored because the World Bank does not publish their data.
Guide tab. Each country has a plain-English guide: weekly news, history and wars, neighbours, money status, challenges, India deals, AI, opportunities and small-money ways to take part. The text is written by hand and kept general where facts were uncertain. Money boxes use the live feeds above. Investment lines are examples, not advice.
Home map. Land is drawn as one silhouette with no country borders, so no disputed boundary is shown. Bubbles sit at approximate country centres.

Why Pakistan is not in the top 50

Pakistan ranks #53 with a score of 39.0, about 1 point below #50. The formula decides this. We did not remove it by hand.

One IMF data series is missing. The IMF World Economic Outlook publishes Pakistan's GDP in US dollars only up to 2025. It gives no dollar projections for 2026-2030 (IMF NGDPD, Pakistan).
So "GDP added 2026-2030" cannot be worked out for Pakistan. The Growth pillar then uses only the IMF growth-rate figure (re-weighted, as for any missing value).
For the same reason, its current GDP uses the latest World Bank figure (2025) instead of an IMF 2026 estimate.
We do not fill the gap with our own guess. If the IMF publishes the series, the next rebuild re-scores Pakistan automatically.

6b. Country data: 200+ sourced parameters

The Data tab on each country dashboard lists every parameter, grouped into Economy, Trade, Debt, People & Welfare, Environment, Security, Democracy, Defence & Space/Tech, Passport and India Relations. Each value is stored with its year and source URL. Missing values stay null and are never estimated. The data is built by scripts/build_params.py (stdlib + openpyxl).

World Bank WDI and Worldwide Governance Indicators (API, most recent non-empty value)
IMF WEO DataMapper: government debt, deficit, inflation, unemployment, current account and more (current year and a 5-year projection)
SIPRI Military Expenditure Database 2026 (published xlsx)
Global Hunger Index 2025 (published xlsx; high-income countries are not scored)
IEP Global Terrorism Index 2026, covering 2025 (published CSV)
V-Dem democracy indices via Our World in Data. The EIU Democracy Index is not used because its data is not redistributable.
Yale Environmental Performance Index 2026 (published xlsx)
Henley Passport Index and Henley Openness Index
WIPO Global Innovation Index 2025 (published xlsx)
UN Comtrade public API: top HS-2 export and import chapters, plus India's bilateral goods trade with each country
UNOOSA objects launched into space (via Our World in Data)

7. Sources

World Bank: World Development Indicators (API: api.worldbank.org/v2)
IMF World Economic Outlook DataMapper (API: imf.org/external/datamapper/api/v1)
Live: GDELT DOC 2.0 API, Google Trends RSS, Yahoo Finance chart API, Moscow Exchange ISS, Google News RSS (fallback only).
Country text: official government and foreign-ministry sites, MEA India, Reuters, BBC and the CFR Global Conflict Tracker, cited inline on each country page.

Indicator	Code	Data years in top 50

8. Limitations

This is a composite model, not a forecast. Different weights give different ranks, so the weights are published for that reason.
Data lags: R&D and military data trail by 1–3 years, and some countries much more (e.g., UAE military spending was last reported for 2014). Stale values are flagged ⚠ on country pages.
No direct measures of governance, institutions, energy security, soft power, nuclear capability or alliances.
Median age is not in the World Bank API, so we use working-age and 65+ shares instead.
Live feeds are rate-limited: GDELT allows about 1 request per 5 s per IP and often returns HTTP 429 from shared IPs, so a full news pass takes several minutes and may fall back to cached or Google News data. Yahoo's endpoint is unofficial and may change. Index levels are delayed.
The attention median is taken over the countries that have fresh GDELT tone in that run. While GDELT is rate-limited, that can be only a handful of countries, and the rest get news_pts = 0.
On weekends and holidays, index changes are frozen at the last session, so index values move only with news tone and the sparklines can be flat.
GDELT matches the country name anywhere in an article, so some stories only mention the country in passing.
Taiwan is excluded because the World Bank WDI does not cover it.
Military spending as % of GDP rewards countries at war or under threat. That is why it carries only 20% of the defence pillar.
"From India's perspective" means the comparisons and commentary are framed for Indian readers. The score itself is country-neutral.

How the Future Power Score and Country Exchange work