Leader containment
Who should pay to slow the leader, and when should you decline to help?
01Estimate the threat
02Share the burden
03Preserve your own route
The decision at the table
Moving the robber onto a leader can benefit every rival while costing you a better theft or production target. Refusing to trade with a leader may be sensible, but a blanket embargo can deny you the only route back into the game.
The mechanism to test
Estimate immediate winning routes, likely hidden potential, and whose turn arrives next. Compare the cost of blocking with your own continuation. Temporary cooperation creates a shared benefit, but the burden can fall unevenly. A threat to punish is credible only if carrying it out remains worthwhile when the time comes.
Proposed experiment
Compare own-resource robber placement, visible-point targeting, and threat-aware targeting that includes a cost of retaliation. Separately test public coordination language with the same actions. Hypothesis: threat-aware containment improves wins without systematically donating the game to the second-place rival.
Use a public timeline of robber moves, lost production, award transfers, and winning opportunities. Analyze the focal player’s win rate and the distribution of other winners. Predefine a diagnostic for avoidable rival wins; motive cannot be recovered from outcome alone.
First ablation
Switching every opponent term off changed the contrast by +0.000 wins per game (95% interval -0.105 to +0.105), and raising them by half through the aggression slider by +0.020 wins per game (95% interval -0.084 to +0.123): against builders that do not retaliate, containment terms neither help nor hurt at depth 2. Details and every run ID are in the ablation report; this is a development-tier result on one lineup.
Endgame racing
Three switches that treat the late game differently from the middle looked strong in one seating and carried no effect once the seating was averaged out. With the candidate always in slot 0, a race leaf against a near-winning leader won by +0.129 wins per game on 64 seeds (+0.029 to +0.228), counting the victory cards inferred from public purchases by +0.172 (+0.080 to +0.264), and biasing the robber and blocking roads toward the leader after two thirds of the game by +0.168 (+0.073 to +0.263). Those are the size of the arena's seating term. Swapped pairs on fresh boards 64 to 127, both halves on one engine, put the seating-corrected effects at +0.018 (−0.018 to +0.053), +0.006 (−0.030 to +0.041), and +0.012 (−0.018 to +0.041), each crossing zero: the switches stay off by default and the strength claim is withdrawn. The endgame report has the mechanisms, all twelve run IDs, and the seating terms; this is development-tier evidence on one lineup of fixed builders.
Robber placement and victim choice
The robber shortlist can target instead of just blocking the most pips. Three switches each beat the default shortlist on 256 paired development boards (blocking the leader's best hex +0.125, stealing the card this seat needs +0.160, sparing trade partners +0.117 wins per game, each interval excluding zero), but none held that strength on fresh seeds, where all three read between +0.035 and +0.039 with intervals crossing zero. The direction is consistent, nothing is confirmed, and the switches stay off by default. The default search lands on the leader's best hex on a quarter of its robber moves; the leader arm raises that to a third. The robber-targeting report has the arms, the plot, and where the robber went.
Third round: under the learned leaf
The four switches came back under the learned n-tuple leaf, each again one switch on top of the tables baseline, now as swapped pairs from the start and through three stages: development on the hand-leaf seeds 0-63, confirmation and population on fresh seeds 800-863, and a preregistered extension to seeds 864-927 for the two arms whose pooled builders effect stayed positive near zero (need and leader-biased endgame). All 36 cohorts completed. Pooled over stages, the seating-corrected effects are leader +0.007 and +0.031, need +0.016 and +0.009, threat −0.005 and +0.043, block leader +0.011 and +0.015 wins per game in the builders and population lineups, every interval crossing zero; the block leader population extension is the only single cohort to exclude zero (+0.033), and combining the two positive arms disagreed across lineups (+0.035 builders, −0.039 population). The mechanisms visibly move the robber, none moves wins, and no switch is worth the browser. The robber-tables report has every run UUID, the hand-leaf comparison, and where the robber actually went.
What could disprove it
Targeting the visually strongest player can miss hidden points. Successful containment is not necessarily a personal success. Keep kingmaking, spite, and rational self-interested blocking as hypotheses rather than labels inferred from one move.
Agent notes
Use the shared experiment design to freeze candidate versions, full lineups, budgets, sample size, primary contrast, and stopping rules before collecting evidence. This is a draft study brief, not a preregistration. No run IDs exist for this proposal.
The deployable policy reads only its own observation and recipient-visible events. The diagram is a conceptual schematic. Build new measured exhibits from retained artifacts using the visual publishing guide.
See the experiment program for fixed tables, opponent classes, and interference between players.
All approaches · Player’s guide · Experiment program
Tracked investigations
| Updated | Investigation | Status | Finding and next step |
|---|---|---|---|
| 2026-09-09 | Maximal aggression | Active | Round one, 256 deterministic games: contrast +0.020 wins per game (95% interval -0.084 to +0.123). See the ablation report. Next: Read the confirmation cohort on fresh seeds where registered; otherwise retest under the confirmed depth-3 candidate before changing defaults. Log 2026-09-09 |
| 2026-09-09 | Opponent progress and production terms | Active | Round one, 256 deterministic games: contrast +0.000 wins per game (95% interval -0.105 to +0.105). See the ablation report. Next: Read the confirmation cohort on fresh seeds where registered; otherwise retest under the confirmed depth-3 candidate before changing defaults. Log 2026-09-09 |
| 2026-09-11 | Endgame racing and stopping the leader | Complete | Withdrawn as a strength claim. The single-seating cohorts (candidate always in slot 0) measured the arena seating term: +0.129, +0.172, and +0.168 wins per game on seeds 0 to 63 and +0.078, +0.039, and +0.051 on seeds 64 to 127, the size of the measured bias. Swapped pairs under containment/endgame-seating put the seating-corrected effects at +0.018 (95% interval -0.018 to +0.053), +0.006 (-0.030 to +0.041), and +0.012 (-0.018 to +0.041), each crossing zero. No switch is confirmed and the defaults stay unchanged. Next: None for the switches as they stand; any future endgame cohort is registered as a swapped pair from the start. Log 2026-09-11 |
| 2026-09-11 | Endgame switches in both seatings | Complete | Swapped pairs on seeds 64 to 127, both halves on the current engine (it no longer reproduces the original confirmations): race leaf +0.018 wins per game (95% interval -0.018 to +0.053), hidden points +0.006 (-0.030 to +0.041), leader bias +0.012 (-0.018 to +0.041), each with an interval crossing zero. The seating terms were +0.092 (+0.014 to +0.169), +0.041 (-0.047 to +0.129), and +0.059 (-0.029 to +0.146). The single-seating contrasts of containment/endgame-race were seating; the corrected effects are consistent with none and exclude anything above about +0.05. Next: No further cohorts for the switches at depth 2 against this lineup; re-open only with a swapped-pair registration, a different lineup, or the deeper confirmed configurations. Log 2026-09-11 |
| 2026-09-12 | Robber placement under the learned leaf | Complete | All four switches are inconclusive under the learned leaf. Tested as swapped pairs through three stages (development seeds 0-63, confirmation and population on seeds 800-863) with a preregistered extension to seeds 864-927 for need and block_leader, 36 cohorts of 256 games all complete, the seating-corrected effects pooled over stages are leader +0.007 (95% interval -0.035 to +0.049) and +0.031 (-0.030 to +0.092), need +0.016 (-0.008 to +0.041) and +0.009 (-0.018 to +0.036), threat -0.005 (-0.042 to +0.033) and +0.043 (-0.005 to +0.091), block_leader +0.011 (-0.005 to +0.027) and +0.015 (-0.007 to +0.036) wins per game, builders and population lineups. The block_leader population extension is the only cohort whose interval excluded zero (+0.033, +0.007 to +0.060) and it does not survive pooling. Combining the two positive arms read +0.035 (-0.015 to +0.085) in the builders lineup and -0.039 (-0.090 to +0.012) in the population lineup. No switch is worth the browser; all stay off. Next: None for these switches in either leaf. The hand-leaf refutation of the need rule does not reproduce under the tables, so any future robber work should name which leaf it prices; block_leader is the only arm to revisit, and only with a different mechanism or a lineup with distinct retaliators. Log 2026-09-11 |
| 2026-09-11 | Robber placement and victim choice | Complete | All three arms beat the default shortlist with the arm in slot 0 on seeds 0-63 (+0.125, +0.160, +0.117) and read near +0.035 on fresh seeds, but swapped pairs under containment/robber-targeting-bias show those gains were the arena seating term: bias-free the effects are leader -0.023, need -0.045 (the only interval excluding zero, on the wrong side), and threat -0.033 wins per game on seeds 64-127. No switch helps, the need rule is refuted at this table, and all three stay off by default. Next: None for these arms. Any future two-search cohort in this arena needs a swapped pair or a null arm; the arms could be revisited against opponents that retaliate, where the leader and threat orderings might matter more than against fixed builders. Log 2026-09-11 |
| 2026-09-11 | Robber targeting against the seating bias | Complete | Three swapped pairs on seeds 64-127 cancelled the arena seating term for the robber-targeting arms: bias-free effects leader -0.023 (95% interval -0.073 to +0.026), need -0.045 (-0.086 to -0.004, the only interval excluding zero), threat -0.033 (-0.084 to +0.018) wins per game, with a seating term of +0.059 to +0.084 for identical searches in adjacent slots in this binary. The slot-0 contrasts of the development cohorts were mostly seating term plus seed selection. Next: Every future two-search cohort in this arena should carry a swapped pair or a null arm; the seating term itself is owned by reports/arena-seating.mdx. Log 2026-09-11 |