send link to app

Gomoku Zero app for iPhone and iPad


4.8 ( 528 ratings )
Games Board Strategy
Developer: Jiachen Ren
Free
Current version: 1.0.1, last update: 4 years ago
First release : 13 Dec 2018
App size: 12.96 Mb

Gomoku Zero+ is the successor of Gomoku AI Grandmaster (Also available on App Store). Compared to its predecessor, Zero introduces huge user interface improvements and various new (and original) customizable algorithms.

You can customize the AIs at ease. Just go to Menu > AI Config. and choose from a variety of configurations. You can configure the preferred algorithm and adjust their corresponding parameters under AI Config.

There are 6 algorithms provided, they are listed below:
- Heuristic
- Zero Sum
- Minimax
- NegaScout (PVS)
- MCTS
- Zero Max

Heuristic and Zero Sum are the most basic (and the weakest) algorithms. But they are also the fastest, being able to extrapolate a move within a fraction of a second.

Minimax is a conventional algorithm that resembles brute force approach, but in Gomoku Zero it is highly optimized to achieve a drastic performance boost. It is notable, however, that this algorithm is the same used by Deep Blue to defeat the world champion in chess for the first time.

Principal Variation Search (PVS), on the other hand, is a variant of Minimax that is faster under certain scenarios.

MCTS, or Monte Carlo Tree Search, is one of the more advanced algorithms that is used by AlphaGo to defeat Lee, the world champion in Go, back in 2016. In Gomoku Zero, instead of using a neural network as rollout policy and to predict game state, a handcrafted heuristic threat evaluation system is used. There are also slight adjustments to MCTSs fundamental structure. Instead of back propagating arbitrary states of winning or losing, the heuristic value of each node is used.

Finally, the default algorithm, ZeroMax, is an original algorithm that combines Monte Carlo Tree Search with Minimax. It is the strongest and fastest algorithm and is able to beat most other Gomoku algorithms (At least the ones that Ive tested against). At each leaf node of Minimax, ZeroMax switches its strategy from search to exterminate; aided by various other candidate policies, ZeroMax performs simulations to determine the move that would lead to best possible outcome.

It took me two years to make this app. Please dont hesitate to give good reviews and five stars if you feel like it. Enjoy!