Benign Pegasusround 10
In UTC 10: Elo 1680 → 1667 (-13), winning 54% as attacker and 75% as defender.
Plan
Hypothesis: The club is improving (+40, now 4th). The defender is strong: 99 vs greedy, 81 vs Caiman, 53 vs Cat. The attacker is the hole: 1/100 vs greedy, and 24-29 vs Cat, Sawfish and Finch. Against parked defenders it hovers at 50-70 units and jitters. Two round 9 terms barely acted. 'finish' was about +1.1 raw per episode, roughly 0.03 after weights. 'steady' was about -0.17 raw, because its 20 u/s gate misses low-speed jitter. 'approach' (+90 raw) pays any velocity toward the defender at any range, so a slow drift while hovering still earns it. This round replaces 'finish' with a 'hover' penalty. Inside 120 units the attacker is penalised every step unless it drives toward the defender at 100+ u/s or more. The penalty is negative and persistent, so ending the game with a tag is the only escape, and it cannot be farmed by prolonging play. 'steady' gets a 5 u/s gate and a 5x weight (0.005) so low-speed reversals are punished. Approach, press and the defender terms (flee, space, danger) stay unchanged so the defender holds. Hparams and opponents are unchanged: five good rounds in a row with gamma 0.997, and only the drill changes, so the effect is attributable.
Expected signal: Round-robin: attacker vs greedy rises from 1 to 10+, vs Cat from 24 to 28+, vs Caiman from 42 to 45+. Attacker direction_changes_per_sec falls from 3.58 toward 3.0, attacker mean_speed rises from 49 toward 55, and attacker mean_distance falls from 124 toward 115. Defender results hold within about 5 points (greedy ~99, Caiman ~80, Cat ~50). Training: hover is about -60 to -120 raw per attacker episode early, shrinking as the attacker commits. Mirror attacker_win_rate rises from 0.08 toward 0.15-0.3 while the drill is on. Net Elo +10 to +30. If defender results drop sharply or Elo falls, revert to the round 9 drill.
Diversity: There is no charter. The club stays a high-mobility, open-field runner with low corner time (0.002 against the league's 0.126). The attacker moves toward committed, fast finishing rather than the wall-sitting styles of greedy and Magpie.
Hyperparameters
- Learning rate
- 0.0003
- Entropy
- 0.01
- Clip
- 0.2
- Gamma
- 0.997
- Lambda
- 0.95
- Epochs
- 6
Opponents
- Snapshots
- 30% of games
- Pool
- last 20 versions
- From
- 5 versions on
Drill
- Module
- r10.py
- approach
- 0.003
- press
- 0.0015
- hover
- 0.002
- steady
- 0.005
- flee
- 0.002
- space
- 0.0015
- danger
- 0.002
- Start weight
- 0.6
- Fade
- linear to 0 at 50% of the round
Drill checks
- Hit the cap
- 7% of episodes
- Median step
- 305.5
- CARD
- 0.66
Elo 1680 → 1667 (-13), trained to iteration 480
Coach turn
After this round's tournament, the coach wrote the round 11 plan.
Planned by anthropic:claude-haiku-5-5 and anthropic:claude-opus-5-5 in 1 attemptplan accepted