j130 learned movement: outcome label (P(hit|state,g)) mode + Gate A; pre-registered outcome arms

ModularBot_garage/src/env_report.nim

@@ -511,7 +511,7 @@ proc knownEnvNames*(): seq[string] =
     LearnedDecayEveryEnv, LearnedDecayShiftEnv, LearnedAlphaEnv,
     LearnedTravelEnv, LearnedReversalEnv, LearnedPrefDistEnv,
     LearnedDistBandEnv, LearnedRadialFracEnv, LearnedWallMarginEnv,
-    LearnedGlobalEnv, LearnedLogEnv,
+    LearnedGlobalEnv, LearnedLabelEnv, LearnedLogEnv,
     # harness vars (read by the test framework, inherited by the bot, so they
     # must NOT be reported as typos)
     "TR_SERVER_JAR", "TR_BATTLE_RUNNER", "TR_BATTLE_RUNNER_DIR",

ModularBot_garage/tests/test_learned_surfer.nim

@@ -76,6 +76,12 @@ proc run(ticks: int, decayShift: int): LearnedSurferModule =
   for _ in 0..<ticks:
     discard result.step(s)
 
+proc runOutcome(ticks: int, decayShift: int): LearnedSurferModule =
+  putEnv(LearnedLabelEnv, "outcome")
+  result = run(ticks, decayShift)
+  putEnv(LearnedLabelEnv, "")
+  loadLearnedEnv()
+
 proc counters(m: LearnedSurferModule): int =
   var n = 0
   for c in m.sbc.counters:
@@ -135,6 +141,31 @@ proc main() =
   let g = run(ticks = 200, decayShift = 1)
   check "TR_LEARNED_GLOBAL still moves (prior-only map)", g.decisions > 20
   delEnv(LearnedGlobalEnv)
+
+  # 8. outcome label: the 2-class counted SBC learns and reads a probability
+  putEnv(LearnedGlobalEnv, "")
+  let o = runOutcome(ticks = 600, decayShift = 1)
+  check "TR_LEARNED_LABEL=outcome still moves", o.decisions > 100
+  check "outcome memory accumulated",
+        (block:
+          var n = 0
+          for c in o.outcome.counters:
+            if c != 0'u8: inc n
+          n > 0)
+  check "outcome hit prior is a probability in [0,1]",
+        (block:
+          let p = o.predictHit(0, 0, 10)
+          p >= 0.0 and p <= 1.0)
+  let o2 = runOutcome(ticks = 300, decayShift = 1)
+  let o3 = runOutcome(ticks = 300, decayShift = 1)
+  check "outcome mode deterministic",
+        (block:
+          var n2, n3 = 0
+          for c in o2.outcome.counters:
+            if c != 0'u8: inc n2
+          for c in o3.outcome.counters:
+            if c != 0'u8: inc n3
+          n2 == n3)
   putEnv(LearnedDecayShiftEnv, "")
   putEnv(LearnedDecayEveryEnv, "")
   loadLearnedEnv()