How many conflict deaths will there be in Sudan in 2025?
Phil’s AI bot has joined Jeremy’s Multi-AI Oracle as side competitors with the machine forecasting bots of the VIEWS competition.
Okay, let’s break down the forecast for conflict deaths in Sudan for the calendar year 2025. The resolution date is January 1, 2026, meaning we are forecasting the total count for the 12 months of 2025. We are currently on April 1, 2025, so the first three months of the year have already passed, and we are forecasting for the remaining nine months plus accounting for the deaths that likely occurred in Jan-Mar 2025.
**1. Understanding the Metric:**
The core task is to estimate “conflict deaths.” Based on standard conflict data collection practices (like ACLED or UCDP), this typically refers to deaths resulting directly from organized violence (combat, bombings, targeted killings by conflict parties). It usually excludes indirect deaths from disease or starvation, even if conflict-induced. Without specific resolution criteria defining the term, I will proceed with this standard definition, but acknowledge this as a key uncertainty. The CSIS source explicitly mentioned high likelihood of “starvation deaths” in 2025, which may or may not be included depending on the unstated fine print.
**2. Analyzing Recent Data and Trends:**
The most crucial piece of information comes from Assistant 2, citing an estimated 29,700 deaths over the past two years (approx. April 2023 – March 2025). This averages to about 14,850 deaths per year, or roughly 3,700 deaths per quarter. This suggests that Jan-Mar 2025 may have already seen deaths in the range of 3,000-4,000.
Other reports confirm ongoing high-intensity conflict:
* An attack killed 7 in Darfur on April 1, 2025.
* A market airstrike in North Darfur killed hundreds recently.
* Evidence of mass graves and torture points to severe violence (Africa Center).
* The Sudanese Armed Forces (SAF) have made recent advances, particularly around Khartoum, while the Rapid Support Forces (RSF) appear to be consolidating in Darfur (CSIS). This shift doesn’t necessarily mean less violence; it could lead to concentrated, intense fighting in Darfur, a region with a history of mass atrocities.
* The humanitarian crisis is catastrophic (UN, SWR Kultur), with millions displaced and facing hunger. While potentially separate from direct conflict deaths, such desperation can fuel further violence.
* International dimensions (ICJ case vs. UAE, US sanctions) suggest external factors may prolong the conflict.
**3. Evaluating Future Prospects (April – December 2025):**
The conflict appears to be entering a new phase, not winding down. The SAF offensive and the RSF’s focus on Darfur suggest continued fighting. Mediation efforts have so far failed to yield a ceasefire. The baseline scenario, based on the past two years’ average rate (~14,850 deaths/year), points strongly towards the highest category (‘More than 12,000’).
Factors favoring continued high or escalating deaths:
* Concentration of fighting in Darfur.
* Ongoing external support for warring parties.
* Lack of effective peace process.
* Deepening humanitarian crisis potentially exacerbating conflict dynamics.
Factors favoring lower deaths:
* Potential for military stalemate leading to reduced large-scale offensives.
* Resource exhaustion on both sides.
* Unforeseen breakthrough in peace negotiations.
* Possibility that the initial death toll estimate (29,700 in 2 years) was an overestimate.
Weighing these factors, a continuation of high levels of violence seems more probable than a significant de-escalation during the remainder of 2025.
**4. Assessing Probabilities for Each Option:**
Based on the average annual rate of ~14,850 deaths, the baseline expectation for 2025 falls into the ‘More than 12,000’ category.
* **Less than 1000 / 1000-3000 / 3000-5000:** Extremely unlikely. The estimated deaths for Jan-Mar 2025 (~3700) already push the total towards the upper end of the 3000-5000 bracket. Achieving these low totals would require an almost immediate and lasting cessation of major hostilities, which seems improbable given current dynamics. These bins get very low probability.
* **5000-8000:** Requires a significant reduction in violence compared to the 2-year average for the rest of the year. Possible if fighting stalemates significantly, but unlikely given the Darfur situation. Moderate-low probability.
* **8000-12000:** Requires the annual rate to drop somewhat below the recent average. Plausible if offensives slow or if the baseline average was slightly high. A significant probability chunk resides here, acknowledging uncertainty.
* **More than 12000:** This aligns with the calculated average annual death rate from the past two years. Given the factors suggesting continued or potentially escalating violence (especially in Darfur), this appears to be the most likely outcome. It receives the highest probability.
**5. Final Probability Distribution (Considering Brier Score & Uncertainty):**
I need to reflect the strong indication towards >12,000 deaths based on the data, while acknowledging the inherent uncertainties in conflict data and future trajectory. Avoiding extreme overconfidence is key for the Brier score. The distribution assigns the highest likelihood to >12,000 but still gives considerable weight (over 30%) to the bins immediately below it.
Less than 1000: 1%
Between 1000 and 3000: 2%
Between 3000 and 5000: 4%
Between 5000 and 8000: 10%
Between 8000 and 12000: 23%
More than 12000: 60%
Pictured above, the main dashboard of the VIEWS machine forecasting competition. Currently, the site is being rebuilt, patience, please.
Background on VIEWS
Near the end of 2024, the latest forecasting competition of the Violence and Impacts Early-Warning System (VIEWS) was launched. On Dec. 4, 2024, during a telecon, a group of machine forecasting experts involved in VIEWS observed Botmaster Jeremy’s demonstration of his Multi-AI Oracle. Although it was too late for Jeremy to sign up as a VIEWS competitor, they urged him to run a side competition for everyone’s benefit. Jeremy chose their Sudan conflict deaths question. Now collaborating with Jeremy, Botmaster Phillip Godzin has joined our side competition with VIEWS. Phil is famed for having won the Metaculus Q4 AI Forecasting Benchmark Tournament
