Overview

Like its predecessor Blockade, the object of Surround is to maneuver a square across the screen, leaving a trail behind. A player wins by forcing the other player to crash into one of the trails. Twelve game variations include options allow for speed-up, diagonal movement, wrap-around, and “erase” (the choice to not draw at a given moment). In addition, the sprites can be set to operate at a slower speed, or progressively speed up through five speeds.

Description from Wikipedia

State of the Art

Human Starts

Result Method Type Score from
7.0 RainbowDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
5.9 PERDDQN (rank) DQN Prioritized Experience Replay
5.5 ApeX DQN DQN Distributed Prioritized Experience Replay
5.4 Human Human Deep Reinforcement Learning with Double Q-learning
4.5 DistributionalDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
4.0 DuelingDDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
1.9 DDQN DQN Deep Reinforcement Learning with Double Q-learning
-0.2 DuelingPERDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
-0.9 PERDDQN (prop) DQN Prioritized Experience Replay
-3.1 NoisyNetDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
-5.3 PERDQN (rank) DQN Prioritized Experience Replay
-6.0 DQN2015 DQN Dueling Network Architectures for Deep Reinforcement Learning
-8.3 A3C LSTM PG Asynchronous Methods for Deep Learning
-9.6 A3C FF (1 day) PG Asynchronous Methods for Deep Learning
-9.7 A3C FF (4 days) PG Asynchronous Methods for Deep Learning
-9.7 Random Random Deep Reinforcement Learning with Double Q-learning
-9001.0 DQN2015 DQN Asynchronous Methods for Deep Learning
-9001.0 GorilaDQN DQN Asynchronous Methods for Deep Learning

No-op Starts

Result Method Type Score from
10.0 NoisyNet-DuelingDQN DQN Noisy Networks for Exploration
9.7 RainbowDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
8.9 PERDDQN (rank) DQN Dueling Network Architectures for Deep Reinforcement Learning
7.1 ApeX DQN DQN Distributed Prioritized Experience Replay
6.8 C51 Misc A Distributional Perspective on Reinforcement Learning
6.5 Human Human Dueling Network Architectures for Deep Reinforcement Learning
6.2 DistributionalDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
4.4 DuelingDDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
2.1 PERDDQN (prop) DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
1.2 DuelingPERDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
1.0 NoisyNet-A3C PG Noisy Networks for Exploration
1.0 DuelingDQN DQN Noisy Networks for Exploration
-1.0 NoisyNet-DQN DQN Noisy Networks for Exploration
-2.5 DDQN+PopArt DQN Learning values across many orders of magnitude
-2.9 DDQN DQN Dueling Network Architectures for Deep Reinforcement Learning
-3.3 NoisyNetDQN DQN Rainbow: Combining Improvements in Deep Reinforcement Learning
-5.6 DQN2015 DQN Dueling Network Architectures for Deep Reinforcement Learning
-6.0 DQN DQN Noisy Networks for Exploration
-8.0 A3C PG Noisy Networks for Exploration
-10.0 Random Random Dueling Network Architectures for Deep Reinforcement Learning