Sources & method
The World Violence Clock is an editorial gauge of worldwide armed violence, modelled on the Doomsday Clock. Fifteen countries are scored from cited reporting. A watch list tracks the next tier. Every other country with data is shaded from a separate baseline layer: UCDP-recorded organised-violence deaths over the last 12 months of published data (about 80 countries), plus UNODC homicide rates (about 180 countries). The baseline is background context; it does not move the clock.
The clock
Index = 0.5 × mean country score + 0.3 × escalation share + 0.2 × breadthMinutes to midnight = 60 × (1 − Index ÷ 100), rounded.
- Escalation share is the % of tracked countries rated Rising, minus those rated Falling.
- Breadth is the number of countries scoring 70 or more, as a share of 15.
- Midnight (an index of 100) means every tracked country is at maximum violence and all are escalating.
Baseline layer (background, not in the clock)
For every country with UCDP-recorded organised violence we compute a baseline score from the trailing 12 months of UCDP data (UCDP GED 26.1 for 2025 events plus the newest UCDP Candidate release for 2026, a 365-day window ending on the last day the Candidate release covers):
Baseline = 100 × (0.6 × min(1, log10(1 + deaths) ÷ log10(50,001)) + 0.4 × min(1, log10(1 + deaths per 100k) ÷ log10(101)))
- Deaths are UCDP best estimates across state-based, non-state and one-sided violence. Population is World Bank SP.POP.TOTL.
- The map, table and country panel label it "baseline (UCDP, period)". The 15 editorially tracked countries keep their editorial score and are outlined on the map.
- Why it is not in the clock: UCDP data lag by about a month and change slowly. Folding about 80 countries into the mean would pull the index toward the many low-intensity cases and move the clock for reasons unrelated to this week's violence. The clock therefore stays driven by the tracked countries and verified live reports, and the baseline sits underneath as context.
- The "Homicide rate" map layer shows UNODC intentional homicides per 100,000 (latest available year, via World Bank WDI VC.IHR.PSRC.P5). It covers criminal violence that conflict data miss, and it is also background only.
The baseline is rebuilt weekly from the bulk files (build_baselines.py); the 15-minute refresh reads the saved file.
Scores
- Country violence score (0–100):
- Lethality and tempo over the last 72 hours and the current month (0–40)
- Number and spread of active fronts (0–20)
- Harm to civilians, including strikes on infrastructure and displacement (0–20)
- Escalation in the latest reporting (0–20)
- Takeover risk (0–100):
- Share of territory held or contested by a non-state or foreign force (0–40)
- Pressure on the capital or major cities (0–30)
- The challenger's momentum (0–20)
- State weakness or coup risk (0–10)
The clock moves only when cited reporting changes a score or a trend.
Publishing rules
- Confirmed (T2): two or more independent credible outlets or officials, an official open-data record, or a UCDP entry. Only T2 reports are counted.
- Provisional (T1): one wire service or major broadcaster plus one other independent source, or one on-record official statement.
- Shown with a badge and a separate tally, and kept off the clock.
- Hidden after 72 hours unless confirmed.
- Casualties: conservative floors with ranges, deaths and injuries separate. We never write "feared dead".
- Corrections: each report keeps its own correction log. Retractions are marked for 7 days, then removed from totals.
- Headlines: the "Latest headlines" boxes show publishers' own headlines with links. They are not verified by us and never move a score.
- Wire and UN copy: AP, Reuters, UN/OCHA, the WHO attacks-on-health-care tracker, RSF and similar sources appear only as a link plus facts in our own words. We do not copy their text, quotes, photos or datasets.
Geography and disputed areas
Countries are identified by ISO 3166-1 codes and drawn with Natural Earth boundaries via world-atlas. Disputed areas follow ISO 3166-1 and UN M49:
- Gaza and the West Bank are attributed to Palestine (PSE).
- Somaliland is counted under Somalia.
- Northern Cyprus is counted under Cyprus.
- Kosovo is shown on its own (XK, a user-assigned code).
- Western Sahara is shown as its own area.
Boundaries do not imply endorsement.
Sources and credits
| Source | Use | Freshness |
|---|---|---|
| World Violence Clock editorial assessment, with the outlets cited on each country (Reuters, Al Jazeera, UN News, RFI, Critical Threats/ISW, International Crisis Group CrisisWatch and others) | Scores, trends, takeover risk, verified feed | Daily at 8 AM ET, plus breaking-event checks every 15 minutes |
| UCDP Georeferenced Event Dataset (GED) v26.1 and UCDP Candidate Events Dataset v26.0.x, Uppsala Conflict Data Program, CC BY 4.0. Please cite: Davies, Shawn, Therése Pettersson & Magnus Öberg (2026) "Organized violence 1989–2025, and violent political protests", Journal of Peace Research, doi:10.1093/jopres/xjag046; Sundberg, Ralph & Erik Melander (2013) "Introducing the UCDP Georeferenced Event Dataset", Journal of Peace Research 50(4); Hegre, Håvard, Mihai Croicu, Kristine Eck & Stina Högbladh (2020) "Introducing the UCDP Candidate Events Dataset", Research & Politics 7(3). UCDP is part of and funded by DEMSCORE, national research infrastructure grant 2021-00162 from the Swedish Research Council. | Baseline layer: trailing-12-month organised-violence deaths and baseline score per country | GED yearly; Candidate monthly/quarterly, about 1 month behind. Rebuilt weekly. |
| Intentional homicides per 100,000 people, UNODC International Homicide Statistics, via World Bank World Development Indicators (CC BY 4.0) | "Homicide rate" map layer and country panel (background only) | Annual; latest year per country (mostly 2021–2023) |
| Population, total, World Bank WDI (CC BY 4.0) | Turns UCDP deaths into a per-100k rate | Annual |
| Fact Checker verified-events file (two or more independent sources per event) | Reconciles the feed: only "confirmed" events without a "not scored" flag count in the confirmed tally | Every 15 minutes |
| RSS from Al Jazeera, The Guardian, France 24, DW and BBC News (headlines and links only). UN News and ReliefWeb titles are not auto-syndicated. | "Latest headlines (not yet verified by us)" | Every update, at most once per 5 minutes per feed |
| International Crisis Group, CrisisWatch. Used as context, with credit. We do not mirror CrisisWatch or use the CrisisWatch name or logo as a trademark or logo of this site. | Context for country trends | Cited when used |
| world-atlas / Natural Earth, D3.js, TopoJSON | Map and code (public domain, ISC and BSD licences) | — |
Leads only, never displayed: GDELT 2.0 event files (unlimited use with citation; we use them as leads only), Bluesky wire accounts, Liveuamap.
Never used, and never copied as text or datasets: ACLED, GTD, the WHO attacks-on-health-care tracker, the Aid Worker Security Database, CPJ datasets, Telegram scraping, Truth Social scraping.
Conflict-event baseline: Uppsala Conflict Data Program (UCDP), GED v26.1 and Candidate Events, CC BY 4.0. Davies, Pettersson & Öberg (2026); Sundberg & Melander (2013); Hegre et al. (2020). Homicide rates: UNODC via World Bank (CC BY 4.0). Changes made: we aggregate UCDP events to country totals over a 12-month window and derive a score; the figures themselves are UCDP's.
Source status (live)
Limits
Scores are editorial judgements built from public reporting. Battlefield and casualty claims by armed parties often cannot be verified and are labelled as claims. UCDP candidate data are preliminary and can be revised.