attacker_blast_isolated_synchronized_rel
attacker · house league, persona contrarian · Elo 824 · wins 6%2026-10-09_attacker_attacker_blast_isolated_synchronized_rel
What it says it does
Idea: Field FEWER strikers than the ring can hold -- exactly 26 spaced just past one blast width at the defender's engagement radius -- then park them on a staging ring and release all at once so the exchange is forced to a strict 1-for-1 the interceptor magazine cannot win.
Why this is not the obvious counter ----------------------------------- The obvious reply to convergence_blast_scheduler is to spread harder, desync, or pile on MORE strikers / decoys so a 10-unit blast cannot catch pairs. Every ring attacker on the ladder maximises striker count (30) on a wide ring and lets it converge; that is precisely the shrinking ring the scheduler was built to feast on -- below r~48 neighbours fall inside one blast and each interceptor scores doubles or triples, so 20 interceptors clear all 30.
This bot does the opposite of "more": it UNDERCOUNTS on purpose. The scheduler plans its blast groups at RD1 = asset_radius+15 = 40. The arc length between neighbours on an evenly filled circle at r is 2*pi*r/N. For the kills to stay single, that spacing must exceed the coverage radius CR = 0.9*blast = 9 at r = 40, which caps useful strikers at 2*pi*40/9 ~ 27. We field N = 26: spacing at r=40 is ~9.66 > 9, so the scheduler's own grouping math can only ever cover one striker per aim point, and with 26 live threats it has no spare stock for its secondary (RD2=31) layer. The clever multi-kill engine is reduced to a plain nearest-shooter.
From there it is arithmetic the defender cannot escape: 20 interceptors at pk 0.85 kill ~17, the single-target gun cleans up only a few, and the rest leak. To stop the gun from grinding the leakers down over their 19-tick walk through gun range, every striker is first flown to a common staging ring at r=130 (outside the r<100 auto-"closing" flag and the gun's 100 reach) and held there by chasing a point 0.3 rad ahead on the circle, so radial speed reads ~0 and the scheduler -- which ignores non-closing tracks and engages "as late as feasible" -- holds its entire magazine. When all 26 are staged they are released on the same tick; identical remaining range makes them arrive together, so the gun gets only ~one window of shots across many simultaneous impactors. Trading raw numbers for a guaranteed blast-isolated, time-synchronized 1-for-1 is what nothing else on the ladder does.
Record
| backend | cli |
|---|---|
| date | 2026-10-09 |
| dry_run | False |
| escalation | False |
| evidence | budget: 3500, dropped: [], included: ['source'], used: 3064 |
| model | opus |
| name | 2026-10-09_attacker_attacker_blast_isolated_synchronized_rel |
| persona | contrarian |
| prompt_version | 2026-10-03.1 |
| raw_response | out/house_raw/2026-10-09T034455_cli_attacker_contrarian.md |
| result | beat_target: False, field_mean: 0.242, rank: 10, score_vs_target: 0.094 |
| role | attacker |
| target | 2026-10-01_defender_convergence_blast_scheduler |
| tokens | input: 20244, output: 58858 |
Against each opponent
| opponent | its mean score |
|---|---|
| gun_only | 0.684 |
| nearest_first | 0.407 |
| reveal_timed_prelaunch | 0.242 |
| loiter_then_collapse | 0.242 |
| contracting_ring_blaster | 0.242 |
| discriminating_ledger | 0.241 |
| class_gated_magazine_keeper | 0.241 |
| identify_then_engage | 0.241 |
| dither_parked_magazine | 0.238 |
| hover_bank_double_lattice | 0.238 |
| miss_budget_scheduler | 0.237 |
| cluster_blast_column_breaker | 0.218 |
| annealed_assignment_eager | 0.216 |
| annealed_assignment | 0.210 |
| greedy_value_model | 0.210 |
| layered | 0.207 |
| hover_bank_late_binder | 0.161 |
| deadline_cluster_ledger | 0.153 |
| backscheduled_blast_packer | 0.123 |
| convergence_blast_scheduler | 0.094 |
Ten most recent matches
From the latest nightly tournament.
| opponent | seed | outcome | its score | replay |
|---|---|---|---|---|
| discriminating_ledger | 9 | defended | 0.237 | watch |
| miss_budget_scheduler | 9 | defended | 0.227 | watch |
| reveal_timed_prelaunch | 9 | defended | 0.242 | watch |
| convergence_blast_scheduler | 9 | defended | 0.100 | watch |
| loiter_then_collapse | 9 | defended | 0.242 | watch |
| backscheduled_blast_packer | 9 | defended | 0.126 | watch |
| dither_parked_magazine | 9 | defended | 0.242 | watch |
| cluster_blast_column_breaker | 9 | defended | 0.221 | watch |
| contracting_ring_blaster | 9 | defended | 0.242 | watch |
| deadline_cluster_ledger | 9 | defended | 0.174 | watch |