Batch cycle-002. Model 0.1.0, graph 0.2.0, parameters 0.1.0. 30 simulations, seeds 1 to 30; probes on 10. Every figure is a distribution over simulations of a structural model with seed parameter values; it shows what the model implies under these settings, not what will happen. Other nations are frozen at their 2026 ratings. No reconstructed historical data is used in this experiment.
Headline
Ethiopia played 30 rounds from 1996 under the scripted method, with scripted events, 30 times. Its capability index moved from 5.96 to a median of 1.74 (−4.22), with 90 per cent of simulations between 1.72 and 1.77 [experiment_objectives].
Starting position
In 1996 the nation ranks 97 on the index against the frozen field, holds 0 of nine domains at Tier 1 and has a tier rank of 100. Its base income is 114 credits a year, calibrated so that its 2026 position is affordable with headroom for investment.
What was decided
A simulation spent a median of 212 credits on moves. The five groups that took most were Transport Infrastructure (build, 25 credits, 11.8 per cent); Land Power (build, 22 credits, 10.4 per cent); Force Generation & Mobilisation (build, 22 credits, 10.4 per cent); Air Power (build, 22 credits, 10.4 per cent); Intelligence & Assessment (build, 20 credits, 9.6 per cent) [experiment_decisions_by_group]. By mode: build 100 per cent [experiment_decisions_by_group]. Concentration of spending (HHI) had a median of 0.084, and the most repeated action appeared in 43 per cent of rounds.
The leading domain by period: rounds 1 to 5, National Security; rounds 6 to 10, Strategic Infrastructure; rounds 11 to 15, Strategic Infrastructure; rounds 16 to 20, Strategic Infrastructure [experiment_spend_by_period].
What the world did
A median simulation saw 7 shocks and 2 upside events. The most damaging per firing were S03 (60 firings in 30 simulations, −47.9 points each); S06 (120 firings in 30 simulations, −41.1 points each); S10 (30 firings in 30 simulations, −21.6 points each) [experiment_shocks]. Upside events fired 60 times across all simulations [experiment_shocks].
The groups that lost most to shocks, per simulation: Trade Facilitation & Logistics −9.74; Fiscal Capacity & Public Finance −8.59; Energy Infrastructure −7.16; Monetary & Currency Capability −6.88 [experiment_shock_damage_by_group].
The edges that carried most change through the graph, per simulation: Fiscal Capacity & Public Finance to Force Generation & Mobilisation (−1.35 points); Fiscal Capacity & Public Finance to Social Infrastructure (−1.14 points); Water, Sanitation & Waste Infrastructure to Agriculture & Food Infrastructure (−0.91 points) [experiment_propagation_by_edge].
Where it ended
At 2026 the median rank is 192 (5th to 95th percentile 192 to 192), the median count of Tier 1 domains 0 (0 to 0) and the median tier rank 183; the resilient score, the mean score sum over simulations, is 129 [experiment_objectives].
Why
Without shocks the index ends +4.828 higher on average, so the shocks cost about 4.828 index points [experiment_probe_attribution].
Holding instead of following this method changes the index by +0.075: that is the method's worth against doing nothing [experiment_probe_attribution].
Removing any single one of the five largest actions changes the index by between +0.005 and +0.006, for example 1997 build Land Power (6 credits) [experiment_probe_attribution]. The largest actions by credits come late, when income is highest, and are rarely the most valuable.
Removing the largest shock of a simulation raises the index by +0.945 on average [experiment_probe_attribution].
The replay probe reproduces the simulations exactly (index change 0.000000), so the attributions are clean [experiment_probe_attribution].
Surprises
saturation_low (warn): 100% of simulations end with a group at 0 that did not start there (median 5 groups).
retired_words (warn): e4-eth/report.md: 'simulations' 1 time(s), first near line 38; use 'simulations'.
The simulation furthest from the median (seed 28) ends −0.03 from it, with 7 shocks and 2 upside events [experiment_outlier_simulations].
Comparison
This is the baseline (hold) experiment, or there is none for this nation in the batch.
Against history: from a reconstructed index of 5.96 the model ends at 1.74, where the 2026 ratings put it at 5.70. Group scores miss by 3.80 points on average, against 0.76 for assuming nothing changed (skill −4.02); the change in each domain has the right sign in 56 per cent of domains. Furthest below history: trade and investment (−5.86); furthest above: critical technology (−1.81) [experiment_against_history]. The starting position is reconstructed, so part of the error is the reconstruction's.