SirStone 186e005a96 fix(training): cap battles at 60 rounds to bound round-end reward scale ...

Round-end reward = cumulative totalScore/50 grows unboundedly with battle
length; long battles (5841 rounds) blew the critic's value scale: vLoss
10-30 during the 60-round cert, 353 at battle-1 round 1, 25745 by round 3871,
policy drift to 0/6 wins. 60-round battles reproduce the certified regime:
bounded value targets, fresh bot process per battle (clears thread state).

2026-08-19 03:32:31 +02:00

..

run.sh

fix(training): cap battles at 60 rounds to bound round-end reward scale

2026-08-19 03:32:31 +02:00

RunTraining.java

fix(botapi): static event queue storage + end-of-battle train wait

2026-08-19 03:12:22 +02:00

training.env

fix(botapi): static event queue storage + end-of-battle train wait

2026-08-19 03:12:22 +02:00