Research

Digital Surface Labs

"Top 100 Arcade Games for OpenArcade ML Data Collection"

Top 100 Arcade Games for OpenArcade ML Data Collection

A curated list of classic single-player arcade games suitable for browser implementation (HTML5 Canvas + JavaScript). Each game is evaluated for player appeal, implementation feasibility, and ML training data value.

Already implemented: Tetris, Flappy Bird, Snake, Breakout, Space Invaders, Pong, Asteroids

Rating Key

  • ML Value (1-5 stars): How valuable the captured input/screen data is for training models. Higher ratings mean richer state spaces, more varied decision-making, and better visual pattern diversity.
  • Complexity: Simple (< 1 week), Medium (1-2 weeks), Hard (2-4 weeks)

Category 1: Puzzle Games

Games with clear state representations, discrete actions, and deep strategic decision-making. These produce the highest quality ML training data because optimal play requires planning and pattern recognition.

# Game Description Input ML Value Complexity
1 Tetris Rotate and place falling blocks to clear lines Arrow keys ★★★★★ Medium
2 2048 Slide numbered tiles to combine and reach 2048 Arrow keys ★★★★★ Simple
3 Minesweeper Uncover tiles while avoiding hidden mines Mouse click ★★★★☆ Simple
4 Sokoban Push crates onto target positions in a warehouse Arrow keys ★★★★★ Simple
5 Bejeweled / Match-3 Swap adjacent gems to match three or more Mouse click ★★★★☆ Medium
6 Columns Arrange falling columns of colored jewels Arrow keys ★★★★☆ Simple
7 Dr. Mario Match colored pills to eliminate viruses Arrow keys ★★★★★ Medium
8 Puyo Puyo Chain-reaction blob matching puzzle Arrow keys ★★★★★ Medium
9 Pipe Dream / Pipe Mania Lay pipe segments before water flows through Mouse click ★★★★☆ Medium
10 Lights Out Toggle lights to turn them all off Mouse click ★★★★☆ Simple
11 Klotski / Sliding Block Slide blocks to move the large block to the exit Mouse drag ★★★★★ Simple
12 Sudoku Fill a 9x9 grid with digits following constraints Mouse + keys ★★★☆☆ Simple
13 Nonogram / Picross Fill grid cells using row/column number clues Mouse click ★★★★☆ Simple
14 Tower of Hanoi Move disk stacks between pegs following rules Mouse click ★★★☆☆ Simple
15 Puzzle Bobble / Bust-a-Move Shoot colored bubbles to match and pop groups Arrow keys ★★★★☆ Medium

Category 2: Shooters (Fixed / Scrolling)

Rich visual environments with continuous aiming decisions. Excellent for training spatial awareness and threat prioritization models.

# Game Description Input ML Value Complexity
16 Space Invaders Shoot descending alien formations Arrow keys ★★★★☆ Simple
17 Galaga Shoot alien swarms with dive-bombing enemies Arrow keys ★★★★☆ Medium
18 Galaxian Defend against swooping alien attackers Arrow keys ★★★★☆ Medium
19 Centipede Shoot a segmented centipede weaving through mushrooms Arrow keys / mouse ★★★★☆ Medium
20 Missile Command Defend cities from incoming ballistic missiles Mouse click ★★★★★ Medium
21 Tempest Shoot enemies climbing out of a geometric tube Arrow keys ★★★★☆ Hard
22 Phoenix Shoot alien birds with a protective shield Arrow keys ★★★☆☆ Medium
23 1942 Vertical scrolling WWII dogfight shooter Arrow keys ★★★★☆ Medium
24 Xevious Vertical scroller with air and ground targets Arrow keys ★★★★☆ Medium
25 Defender Side-scrolling rescue mission with wraparound world Arrow keys ★★★★★ Hard
26 Robotron 2084 Twin-stick survival shooter, rescue humans Arrow keys + WASD ★★★★★ Medium
27 Sinistar Fly, mine, and destroy the menacing Sinistar Arrow keys ★★★★☆ Hard
28 Raiden Vertical scrolling military shoot-em-up Arrow keys ★★★★☆ Medium
29 Gradius Side-scrolling shooter with power-up system Arrow keys ★★★★☆ Hard
30 R-Type Side-scrolling shooter with charge beam and force pod Arrow keys ★★★★☆ Hard

Category 3: Maze / Navigation Games

Discrete decision-making in structured environments. Ideal for reinforcement learning research due to clear state-action-reward structure.

# Game Description Input ML Value Complexity
31 Pac-Man Eat all dots while avoiding ghosts in a maze Arrow keys ★★★★★ Medium
32 Ms. Pac-Man Pac-Man with randomized ghost behavior and new mazes Arrow keys ★★★★★ Medium
33 Dig Dug Dig through dirt and inflate underground enemies Arrow keys ★★★★☆ Medium
34 Bomberman Place bombs to destroy blocks and enemies in a maze Arrow keys ★★★★★ Medium
35 Lode Runner Collect gold while trapping enemies in dug holes Arrow keys ★★★★★ Medium
36 Boulder Dash Navigate caves collecting diamonds, avoid falling rocks Arrow keys ★★★★★ Medium
37 Q*bert Hop on cubes to change their color while dodging enemies Arrow keys ★★★★☆ Medium
38 Amidar Traverse grid edges to fill in rectangles Arrow keys ★★★★☆ Medium
39 Rally-X Drive through a maze collecting flags, avoid enemies Arrow keys ★★★★☆ Medium
40 Mr. Do! Dig tunnels, collect cherries, crush enemies with apples Arrow keys ★★★★☆ Medium

Category 4: Platformers

Rich continuous-action spaces with timing and spatial reasoning. High visual complexity produces diverse training frames.

# Game Description Input ML Value Complexity
41 Donkey Kong Climb girders and ladders to rescue the princess Arrow keys ★★★★★ Medium
42 Super Mario Bros (simplified) Side-scrolling platformer with enemies and power-ups Arrow keys ★★★★★ Hard
43 Doodle Jump Bounce endlessly upward on platforms Arrow keys / tilt ★★★★☆ Simple
44 Ice Climber Climb ice mountain, break blocks overhead Arrow keys ★★★★☆ Medium
45 Bubble Bobble Trap enemies in bubbles and pop them Arrow keys ★★★★☆ Medium
46 Mega Man (simplified) Run-and-gun platformer with boss fights Arrow keys ★★★★★ Hard
47 Elevator Action Descend through a building using elevators, shoot spies Arrow keys ★★★★☆ Medium
48 Mappy Bounce on trampolines to collect stolen goods, avoid cats Arrow keys ★★★★☆ Medium
49 Snow Bros Roll snowballs to knock enemies off platforms Arrow keys ★★★★☆ Medium
50 Spelunky (simplified) Procedurally generated cave platformer Arrow keys ★★★★★ Hard

Category 5: Action / Reflex Games

Fast-paced gameplay producing high-frequency input data. Great for training reaction-time and prediction models.

# Game Description Input ML Value Complexity
51 Flappy Bird Tap to navigate a bird through pipe gaps Space / tap ★★★★☆ Simple
52 Frogger Guide a frog across traffic and river hazards Arrow keys ★★★★★ Medium
53 Crossy Road (simplified) Endless Frogger-style road crossing Arrow keys / tap ★★★★☆ Simple
54 Fruit Ninja Slash flying fruit, avoid bombs Mouse swipe ★★★★☆ Medium
55 Whack-a-Mole Hit moles as they pop up from holes Mouse click ★★★☆☆ Simple
56 Tapper Serve drinks to advancing customers at a bar Arrow keys ★★★★☆ Medium
57 Kaboom! Catch falling bombs in buckets Mouse ★★★★☆ Simple
58 Joust Flap to fly and lance enemy knights on ostriches Arrow keys ★★★★☆ Medium
59 BurgerTime Walk over burger ingredients to drop them, avoid enemies Arrow keys ★★★★☆ Medium
60 Paperboy Deliver newspapers while dodging suburban hazards Arrow keys ★★★★☆ Hard

Category 6: Racing / Driving

Continuous control with strong visual flow. Produces smooth, high-frequency input streams ideal for imitation learning.

# Game Description Input ML Value Complexity
61 Pong Paddle vs. paddle bouncing ball game Arrow keys ★★★★☆ Simple
62 Road Fighter Vertical scrolling race, dodge traffic and obstacles Arrow keys ★★★★☆ Medium
63 Outrun (simplified) Pseudo-3D highway racing with forks Arrow keys ★★★★★ Hard
64 Spy Hunter Vertical scrolling driving with weapons Arrow keys ★★★★☆ Hard
65 Night Driver First-person night driving on a winding road Arrow keys ★★★★☆ Medium
66 Slot Racer / Slot Car Keep car on track through curves at varying speed Arrow keys ★★★☆☆ Simple
67 Super Sprint Top-down racing around tracks Arrow keys ★★★★☆ Medium
68 Moon Patrol Drive a lunar buggy, jump craters and shoot obstacles Arrow keys ★★★★☆ Medium

Category 7: Snake / Growth Games

Simple state spaces that scale in difficulty. Excellent for baseline ML experiments.

# Game Description Input ML Value Complexity
69 Snake Grow a snake by eating food, avoid walls and yourself Arrow keys ★★★★★ Simple
70 Tron / Light Cycles Leave a trail, force opponent into walls Arrow keys ★★★★★ Simple
71 Nibbler Snake in a closed maze with timed level exits Arrow keys ★★★★☆ Simple
72 Slither.io (simplified) Grow a snake, consume pellets, avoid other snakes Mouse ★★★★☆ Medium
73 Agar.io (simplified) Grow a cell by absorbing smaller cells Mouse ★★★★☆ Medium

Category 8: Brick Breaker / Ball Games

Continuous mouse/paddle control with physics. Great for trajectory prediction training data.

# Game Description Input ML Value Complexity
74 Breakout Bounce ball off paddle to destroy bricks Arrow keys / mouse ★★★★☆ Simple
75 Arkanoid Breakout with power-ups and varied brick layouts Arrow keys / mouse ★★★★★ Medium
76 DX-Ball Enhanced Breakout with many power-up types Mouse ★★★★☆ Medium
77 Peggle (simplified) Aim and shoot a ball to hit orange pegs Mouse click ★★★★★ Medium
78 Pinball (simplified) Flipper-based ball physics game Arrow keys ★★★★☆ Hard
79 Air Hockey Paddle vs. puck physics on a table Mouse ★★★★☆ Simple

Category 9: Tower Defense / Strategy

Complex decision-making over time. Rich state spaces produce high-value training data for planning models.

# Game Description Input ML Value Complexity
80 Plants vs. Zombies (simplified) Place plant defenders to stop zombie waves Mouse click ★★★★★ Hard
81 Tower Defense Classic Place towers along a path to stop enemy waves Mouse click ★★★★★ Medium
82 Rampart Build castle walls, place cannons, fire at ships Mouse click ★★★★★ Hard
83 Lemmings (simplified) Assign jobs to lemmings to guide them to the exit Mouse click ★★★★★ Hard
84 SimCity (micro) Tiny city-builder with budget and zoning Mouse click ★★★★☆ Hard

Category 10: Endless Runners / Dodgers

High-engagement, session-based games. Simple input but continuous decision-making ideal for imitation learning.

# Game Description Input ML Value Complexity
85 Canabalt One-button endless rooftop runner Space / tap ★★★★☆ Simple
86 Geometry Dash (simplified) Rhythm-based jump-and-fly obstacle course Space / tap ★★★★☆ Medium
87 Jetpack Joyride (simplified) Fly with a jetpack, dodge obstacles, collect coins Space / tap ★★★★☆ Medium
88 Temple Run (2D simplified) Endless runner with lane-switching and jumping Arrow keys ★★★★☆ Medium
89 Helicopter Game Hold to ascend, release to descend through a cave Mouse / space ★★★★☆ Simple
90 Chrome Dino Jump and duck over cacti and pterodactyls Arrow keys ★★★★☆ Simple

Category 11: Classic Arcade Miscellaneous

Iconic games that don't fit neatly into other categories but have strong nostalgia appeal and interesting ML properties.

# Game Description Input ML Value Complexity
91 Asteroids Rotate and thrust through space, shoot asteroids Arrow keys ★★★★★ Medium
92 Lunar Lander Control thrust to softly land on the moon Arrow keys ★★★★★ Simple
93 Space Duel Asteroids variant with tethered dual ships Arrow keys ★★★★☆ Medium
94 Battlezone First-person wireframe tank combat Arrow keys ★★★★☆ Hard
95 Warlords Four-corner Breakout-style castle defense Arrow keys / mouse ★★★★☆ Medium
96 Tanks / Artillery Angle and power artillery shots at opponent Arrow keys ★★★★★ Simple
97 Duck Hunt (adapted) Shoot ducks flying across the screen Mouse click ★★★★☆ Simple
98 Simon Repeat increasingly long color/sound patterns Mouse click ★★★☆☆ Simple
99 Brick Breaker Roguelike Breakout with roguelike power-up choices between levels Mouse + click ★★★★★ Medium
100 Cookie Clicker (minimal) Click to earn, buy upgrades, automate Mouse click ★★★☆☆ Simple

ML Data Collection Notes

Highest ML Value Games (5 stars)

These games produce the richest training data due to complex state spaces, diverse strategies, and clear reward signals:

  1. Tetris -- Rich combinatorial state, planning horizon, multiple valid strategies
  2. 2048 -- Pure strategy with complete information, compact state representation
  3. Sokoban -- Planning and search, well-studied in AI research
  4. Pac-Man / Ms. Pac-Man -- Benchmark for RL research (Atari), multi-objective decision-making
  5. Missile Command -- Continuous targeting, prioritization under pressure
  6. Robotron 2084 -- Dual-stick continuous control, crowd navigation
  7. Donkey Kong -- Timing, pathfinding, and threat avoidance
  8. Frogger -- Multi-lane timing and planning, classic RL benchmark
  9. Snake -- Growing constraint space, self-avoidance planning
  10. Tron / Light Cycles -- Spatial reasoning and territory control
  11. Lunar Lander -- Continuous control, classic control theory problem
  12. Lemmings -- Assignment planning, resource management
  13. Boulder Dash -- Physics-based puzzle with real-time pressure
  14. Bomberman -- Spatial reasoning, trap setting, escape planning
  15. Tower Defense -- Placement optimization, resource allocation
  16. Lode Runner -- Strategic trap-setting and pathfinding

Input Types Distribution

Input Type Count Notes
Arrow keys 62 Most common, discrete directional input
Mouse click 18 Positional targeting, good for spatial models
Mouse movement 8 Continuous positional data, high frequency
Space / tap 8 Binary input, timing-focused
Arrow keys + WASD 2 Twin-stick, dual-axis control
Mouse drag/swipe 2 Gesture-based input

Tier 1 -- Implement Next (high player appeal + high ML value + low complexity): - 2048 (Simple, 5-star ML) - Pac-Man (Medium, 5-star ML) - Frogger (Medium, 5-star ML) - Doodle Jump (Simple, 4-star ML) - Lunar Lander (Simple, 5-star ML) - Tron / Light Cycles (Simple, 5-star ML) - Chrome Dino (Simple, 4-star ML)

Tier 2 -- Strong candidates (high appeal, medium complexity): - Bomberman, Galaga, Centipede, Donkey Kong, Missile Command - Arkanoid, Puzzle Bobble, Dr. Mario, Bubble Bobble - Minesweeper, Sokoban, Match-3

Tier 3 -- Worth building but more work: - Tower Defense, Lemmings, Mega Man, Gradius, Outrun - Super Mario (simplified), Spelunky (simplified) - Plants vs. Zombies, Geometry Dash

Data Collection Strategy Per Game Type

Game Type Frame Rate Key Data Points
Puzzle (Tetris, 2048) 5-10 fps Board state, piece placement, score
Shooter (Galaga, 1942) 15-30 fps Player position, enemy positions, bullet trajectories
Platformer (DK, Mario) 15-30 fps Player velocity, platform positions, enemy locations
Racer (Outrun, Road Fighter) 15-30 fps Speed, lane position, obstacle distances
Endless runner (Canabalt, Dino) 10-15 fps Obstacle timing, jump/duck decisions
Mouse-based (Missile Command) 15-30 fps Cursor position, click timing, target selection