Batch mcts-m2. Model 0.1.0, graph 0.2.0, parameters 0.1.0. 50 simulations, seeds 1 to 50; probes on 5. 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
Estonia played 20 rounds from 2026 under the tree search (mcts) method, with stochastic events, 50 times. Its capability index moved from 11.49 to a median of 12.80 (+1.32), with 90 per cent of simulations between 10.84 and 13.88 [experiment_objectives].
Starting position
In 2026 the nation ranks 33 on the index against the frozen field, holds 0 of nine domains at Tier 1 and has a tier rank of 27. Its base income is 347 credits a year, calibrated so that its 2026 position is affordable with headroom for investment.
What was decided
A simulation spent a median of 1,899 credits on actions. The five groups that took most were Semiconductors & Advanced Computing (buy, 150 credits, 7.9 per cent); Space Technologies (partner, 130 credits, 6.9 per cent); Economic Output & Productivity (buy, 99 credits, 5.2 per cent); Energy Technologies (buy, 80 credits, 4.2 per cent); Development Cooperation & Global Assistance (buy, 76 credits, 4.0 per cent) [experiment_decisions_by_group]. By mode: buy 41 per cent, partner 33 per cent, build 26 per cent [experiment_decisions_by_group]. Concentration of spending (HHI) had a median of 0.042, and the most repeated action appeared in 40 per cent of rounds.
The leading domain by period: rounds 1 to 5, Critical Technology; rounds 6 to 10, Critical Technology; rounds 11 to 15, Critical Technology; rounds 16 to 20, Critical Technology [experiment_spend_by_period].
The method's chosen action was within 5 per cent of its best alternative in 100 per cent of actions and ahead by more than 25 per cent in 0 per cent [experiment_decision_margins].
What the world did
A median simulation saw 3 shocks and 1 upside events. The most damaging per firing were S09 (5 firings in 5 simulations, −39.3 points each); S03 (19 firings in 18 simulations, −38.7 points each); S04 (29 firings in 23 simulations, −37.4 points each) [experiment_shocks]. Upside events fired 44 times across all simulations [experiment_shocks].
The groups that lost most to shocks, per simulation: Information Environment & Strategic Communication −1.88; Energy Infrastructure −1.70; Maritime Power −1.67; Trade Facilitation & Logistics −1.65 [experiment_shock_damage_by_group].
The edges that carried most change through the graph, per simulation: Semiconductors & Advanced Computing to Artificial Intelligence (+3.45 points); Space Technologies to Space Operations (+1.54 points); Semiconductors & Advanced Computing to Economic Leverage & Dependence (+1.32 points) [experiment_propagation_by_edge].
Where it ended
At 2046 the median rank is 24 (5th to 95th percentile 19 to 39), the median count of Tier 1 domains 1 (0 to 3) and the median tier rank 16; the resilient score, the mean score sum over simulations, is 906 [experiment_objectives].
Tier 1 probability changed most in: human capital (0 to 80 per cent of simulations); production and innovation (0 to 12 per cent of simulations) [experiment_tiers_by_round].
Why
Without shocks the index ends +1.070 higher on average, so the shocks cost about 1.070 index points [experiment_probe_attribution].
Holding instead of following this method changes the index by −2.333: 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 −0.021 and −0.014, for example 2039 build Advanced Materials & Manufacturing (21 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 +0.638 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
tier_jump (warn): largest gain in one domain within five years: 6 tiers (36 simulations above 2); tiers are reported, not an objective.
The simulation furthest from the median (seed 33) ends −2.84 from it, with 8 shocks and 0 upside events [experiment_outlier_simulations].
Comparison
This is the baseline (hold) experiment, or there is none for this nation in the batch.
No backcast comparison: this experiment does not start from a historical position [experiment_against_history].
The search
Monte Carlo tree search 0.1.0 with candidate generators 0.1.0 (marginal return, graph leverage, goal gap; top 4 groups each), 500 iterations a decision, UCT c = 1.4 scaled to the reward spread, widening alpha = 0.5, greedy playouts to the horizon, objective score_sum; playouts drew their futures from the stochastic scenario.
The committed decision most often chosen, round by round: round 1 build Fiscal Capacity & Public Finance (10 per cent of simulations, visit share 10 per cent); round 2 build Nuclear Deterrence (8 per cent of simulations, visit share 10 per cent); round 3 build Fiscal Capacity & Public Finance (12 per cent of simulations, visit share 10 per cent); round 4 build Fiscal Capacity & Public Finance (16 per cent of simulations, visit share 10 per cent); round 5 buy Economic Output & Productivity (12 per cent of simulations, visit share 10 per cent) [mcts_decision_sequence]. The search hesitated (top two root candidates within 10 per cent) in 44 per cent of decisions, considered a mean of 32 candidates, grew trees of a mean 450 nodes to depth 4.5, and took 1.0 seconds a decision [mcts_decision_sequence].
Where the committed actions came from, by proposing generator (committed simulation-rounds): marginal return 30.7; goal gap 4.7; graph leverage | marginal return + graph leverage 2.9; graph leverage 2.5; marginal return + graph leverage | graph leverage 1.2 [mcts_root_candidates].
Against the greedy experiment with the same seeds (m2-est-greedy-marginal), the search's score_sum at the horizon is −2.470 on average (10th to 90th percentile −5.216 to +1.615), better in 18 per cent of seeds; on the index −0.037, the score sum −2.5 and Tier 1 domains 0.00 [mcts_regret].
Across seeds the same root decision won in a mean 11 per cent of simulations (first five rounds 12 per cent) [mcts_stability].
The principal variation was followed 100 per cent at depth 1, 5 per cent at depth 2, 5 per cent at depth 3, 4 per cent at depth 4 [mcts_pv_vs_path].