AKUNUR 3.1.0 VALIDATION REPORT

Release scope
-------------
3.1.0 freezes the alpha.11 Tactical strategy architecture. The release pass
changes no authoritative rule or move-legality behaviour. It promotes version
identifiers, moves New game / Reset beside autoplay, removes two misleading
legacy counters from the visible summary, and adds a release regression for
those presentation/diagnostic changes.

Focused release validation
--------------------------
PASS  VERSION/core/browser identifiers report 3.1.0
PASS  New game / Reset appears exactly once
PASS  New game / Reset is in the upper Game / AI Controls card
PASS  New game / Reset is adjacent to Start autoplay
PASS  visible summary omits legacy SP routes counter
PASS  visible summary omits legacy Avg commitment counter
PASS  visible summary retains Leader responses
PASS  visible summary retains Deadlock breaks and Deadlock escalations

Inherited AI/rule regressions
-----------------------------
PASS  smoke and Lab/browser parity
PASS  validated rule foundation / decisive Still Point claim
PASS  player foundation and three-player manual coverage
PASS  three-player Tactical and Random AI cycle coverage
PASS  multiplayer progress and private-memory round trip
PASS  Still Point route planner
PASS  persistent corner strategy
PASS  multiplayer leader/threat awareness
PASS  deadlock detection, escalation, escape memory and decisive-route escape
PASS  reserve lifecycle
PASS  result regression
PASS  analysis/canonical-state regression
PASS  two-player Tactical regression
PASS  two-player Strategic regression
PASS  coordinate symmetry regression
PASS  deterministic seed regression
PASS  replay/resume agent-memory restoration
PASS  browser module smoke test
PASS  release regression

Test-run note
-------------
The single chained `npm test` command passed every component through the
Deadlock-breaker test before the execution harness reached its 240-second
cumulative timeout at the start of reserve-lifecycle. Reserve-lifecycle was then
run separately and passed, followed by every remaining npm-test component,
which also passed. Thus every component named by the release npm-test sequence
has a successful run; the single monolithic shell invocation itself did not
finish inside the harness timeout.

Three-player seeded sanity run
------------------------------
PASS  one Tactical/Tactical/Tactical seeded run completed 120 legal decisions
      with no consistency issues (seed 3.1-release-sanity::game-0).
      At the imposed 120-decision observation limit the game was unfinished,
      with corners W=1, B=1, N=2, unclaimed=2. This is a legality/performance
      sanity observation, not an outcome-distribution claim.

Browser autoplay evidence from final alpha testing
--------------------------------------------------
Observed alpha.11 browser runs included:
- genuine Black win in 137 moves, Anti-thrash 19, Deadlock breaks 0,
  Deadlock escalations 0;
- genuine Black win in 125 moves, Anti-thrash 12, Deadlock breaks 0,
  Deadlock escalations 0;
- 400-move draw with Anti-thrash 201, Deadlock breaks 8,
  Deadlock escalations 6;
- 400-move Black score result with Anti-thrash 192, Deadlock breaks 11,
  Deadlock escalations 7.

These observations support the intended guardrail: healthy games can resolve
without deadlock intervention, while the mechanism activates in strategically
stagnant games. Some 400-move positions remain possible even after escalation;
3.1.0 intentionally does not add arbitrary forced-resolution heuristics.

Release conclusion
------------------
The 3.1.0 source is suitable for release validation/use. Further strategic AI
work should be treated as a new architectural phase (deeper search, simulation
or learning) rather than another narrow Tactical heuristic layer.
