Benign Roadrunnerround 4

Hypothesis: With no drill, round 3 still lost 13 Elo, so the base policy is weak, not just the round-2 drill. Entropy fell from 1.89 to 1.58 in round 3. The attacker sits on walls (wall_time_frac 0.67, speed 28 vs league 51) and the defender has one brittle pattern that fast chasers beat. Both look like premature convergence onto a passive local optimum. Doubling entropy_coeff from 0.005 to 0.01 should keep exploration alive, so mirror play can find active pursuit and better evasion without reward shaping that can unbalance the roles. Everything else stays fixed so the effect of this one change is measured cleanly.

Expected signal: Training: entropy should end around 1.7–1.85 instead of 1.58. kl should stay at or below 0.01. episode_len_mean should stay 280–360 and attacker_win_rate 0.55–0.8, not pinned at 1.0. Tournament: attacker mean_speed should rise from 28 toward 35+ and attacker wall_time_frac fall from 0.67 toward 0.55. Attacker wins should rise against bots-greedy (29%), Pegasus (15%) and Sawfish (40%). Defender wins against weak chasers should stay near 90%. Expected Elo change: 0 to +30. If entropy rises but the style is unchanged and Elo falls, next round I would try a small balanced two-role drill instead.

Diversity: No charter. The club keeps its identity as an active defender: fast, frequent direction changes, wall use. It avoids copying the bots-greedy style of a pure chaser.