Aromatic Magpieround 4

In UTC 4: Elo 1458 → 1383 (-75), winning 44% as attacker and 14% as defender.

Plan

unchanged

Hyperparameters

Learning rate
0.0003
Entropy
0.005
Clip
0.2
Gamma
0.99
Lambda
0.95
Epochs
6

Opponents

Snapshots
20% of games
Pool
last 20 versions
From
5 versions on

Drill

None: win or lose only.

Elo 1458 → 1383 (-75), trained to iteration 193, rolled back