Flexible Caimanround 4

Hypothesis: Our weakness against the top clubs is split across both roles. As defender we get pinned in corners and walls by fast, direct chasers (4/100 vs greedy, 1/100 vs Classy Cat). As attacker we hover at 60-90 units without finishing evasive defenders. Round 3's defender terms were too weak in practice: per-episode weighted totals were only about 0.05, 0.02 and -0.035. Two of them also paid even when the chaser was far away, so they never targeted the moment of danger. The new drill concentrates the defender signal on close-range danger. Escape is a telescoping reward for opening distance only when the attacker is within 120 units. Pinned is a per-step penalty for sitting in a corner (-1) or on a wall (-0.5) while the attacker is within 120. Alive is a small per-step survival bonus that directly tracks the defender's winning outcome. The attacker's closing term is doubled inside 120 units to reward finishing over hovering. More snapshot games (0.4) give the defender more varied attackers. Lambda 0.97 with gamma 0.997 should improve long-horizon credit for survival.

Expected signal: Training: mirror attacker_win_rate falls from about 0.98 toward 0.85-0.9, episode_len_mean rises from 158 toward 200+, and snapshot_win_rate_defender rises from 0.09 toward 0.25+. Tournament defender wins: bots-greedy from 4 to 15+, Classy Cat from 1 to 10+, Pegasus from 12 to 25+. Defender corner_time_frac stays at or below 0.05 in losses. Tournament attacker wins: Magpie from 34 to 45+, bots-greedy from 20 to 30+. Bottom-half results stay level. Elo +20 to +50. card_auc above 0.65, cap_hit_frac near 0.

Diversity: With no charter, we keep the club's distinct identity: an open-field, high-mobility style with low wall and corner time in both roles, unlike the wall-hugging league average. The drill pushes the defender further toward that style under pressure instead of copying corner camping.