Finicky Sawfishround 11

Hypothesis: The round-6 regime (gamma 0.997, lambda 0.97, lr 2e-4, entropy 0.01, snapshot 0.5 with pool 20, def_opening drill fading by 50%) has gained Elo five rounds running (+48, +55, +60, +74, +36). The club is now 2nd at 1730. Attacker results improved sharply vs Caiman (30→65) and Classy Cat (18→48). But attacker vs bots-greedy collapsed from 31% to 10%, which met round 10's trigger. The analyst found these are timeouts where our attacker stalls or hovers while greedy parks in a corner. The mirror rarely shows a parked defender, so the attacker never learns to dig one out. I keep hparams, opponents and the def_opening term, so the defender's gains carry over. I add a small telescoping attacker closing term: each step's reduction in distance to the defender, divided by max_dist. Because it telescopes, a game's total is bounded by the starting distance over max_dist, so it cannot saturate the cap. It rewards actually closing, not speed or direction changes, so it should not bring back the jitter of round 2's attacker_stall term. It fades by 50%, leaving match prep on outcome only.

Expected signal: Drill checks: card_auc above 0.6 and cap_hit_frac near 0. Training: - Mirror attacker_win_rate stays in the 0.3-0.65 band. - episode_len_mean stays above 250. - Entropy stays at or above about 0.7. - vf_explained_var stays at or above 0.9. Tournament: - Attacker vs bots-greedy: 10% → 25%+ - Attacker vs Finch: 31% → 35% - Attacker vs Classy Cat: holds around 45-50% - Defender results hold within about 5 points (Cat 53, Magpie 53, Finch 79, greedy 99). - Attacker direction_changes_per_sec stays around 2-2.5. Elo -10 to +30, no rollback. If greedy attacker wins don't move, the stall is a greedy-specific geometry problem: next round, target corner pursuit directly. If defender results drop, remove the attacker term.

Diversity: No charter. The club stays a fast, straight-line chaser (speed 79 vs league 57, few direction changes) and a long-surviving runner that avoids corners. The new term only rewards closing distance, so it reinforces our direct-pursuit identity rather than copying Finch's open-field runner style.