Quoridor

Title: Quoridor
Author: Mirko Marchesi
License: Commercial Board Game

Author: Martijn van Steenbergen
License: Creative Commons
Website: http://martijn.van.steenbergen.nl/projects/quoridor/

It has been a while since I did an entry on an abstract strategy game. So, here is a short one on a different two to four person game call Quoridor. I am only going to focus on the two person version.

The rules are very simple. Quoridor is played on a 9 by 9 board. The goal is to get your marker to other side of the board before your opponent does. The twist is that players can place 2-long barriers on the board. Each turn, you can either move your marker one square orthogonally or place a barrier. But you cannot move through a barrier, neither the opponent's barriers nor your own. You have a limited supply (ten) of barriers; use them wisely. There is a restriction on placing the barriers; you cannot place one which completely blocks your opponent from the other side. One final rule: if your marker is next to your opponent, you can jump over your opponent and land on any neighboring square not blocked by a barrier.

What makes Quoridor interesting is learning what makes for a good position. In A.I. lingo, learning a good heuristic. I really enjoyed the learning possess for this game as the obivious heuristics are flawed. First off, being close to the other side of the board is not always the best place to be. If your opponent has a bunch of barriers left in their pool, they can use them and force you backtrack or zig-zag. Next, I thought that having multiple paths to the other side was a good thing. But oddly, due to the rule that your opponent may not cut off all paths, sometimes having only one path is an advantage as your opponent cannot take it away. The other factor is the value of the barriers, the more in reserve the better, but what is the right weighting of these factors?


To start learning the game, you can play against Martijn van Steenbergen java implementation. It has a very clean and intuitive interface. There are four levels of play. The A.I. is not strong, but the SmartBrain 2 level beat me the first time I played due to my lack of familarity with the game. Level 3 plays a better game, but is slower. Level 4 is too slow to be practical. I would say that the A.I. underestimates the value of keeping barriers in its pool.

One observation I have is that the play balancing of Quoridor is spot on. The choice of using a 9x9 board and a pool of 10 barriers seems to be just right. Reading the wikipedia article on Quoridor mentions some previous incarnations of this sort of game. So, these choices are the result of some trial and error. One of those previous games is Blockage. Another vaguely similar game is Camelot, for which a Zillions of Games zrf file is available.