15.12.2012 Views

Bayesian Programming and Learning for Multi-Player Video Games ...

Bayesian Programming and Learning for Multi-Player Video Games ...

Bayesian Programming and Learning for Multi-Player Video Games ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Chapter 7<br />

Strategy<br />

Strategy without tactics is the slowest route to victory. Tactics without strategy is<br />

the noise be<strong>for</strong>e defeat.<br />

All men can see these tactics whereby I conquer, but what none can see is the<br />

strategy out of which victory is evolved.<br />

What is of supreme importance in war is to attack the enemy’s strategy.<br />

Sun Tzu (The Art of War, 476-221 BC)<br />

We present our solutions to some of the problems raised at the strategic level. The main idea<br />

is to reduce the complexity encoding all possible variations of strategies to a few strong<br />

indicators: the build tree* (closely related to the tech tree*) <strong>and</strong> canonical army compositions.<br />

We start by explaining what we consider that belongs to strategic thinking, <strong>and</strong> related work.<br />

We then describe the in<strong>for</strong>mation that we will use <strong>and</strong> the decisions that can be taken. As we try<br />

<strong>and</strong> abstract early game strategies to “openings” (as in Chess), we will present how we labeled a<br />

dataset of games with openings. Then, we present the <strong>Bayesian</strong> model <strong>for</strong> build tree prediction<br />

(from partial observations), followed by its augmented version able to predict the opponent’s<br />

opening. Both models were evaluated in prediction dataset of skilled players. Finally we explain<br />

our work on army composition adaptation (to the opponent’s army).<br />

Build trees estimation was published at the Annual Conference on Artificial Intelligence <strong>and</strong><br />

Interactive Digital Entertainment (AAAI AIIDE) 2011 in Palo Alto [Synnaeve <strong>and</strong> Bessière,<br />

2011] <strong>and</strong> openings prediction was published at Computational Intelligence in <strong>Games</strong> (IEEE<br />

CIG) 2011 in Seoul [Synnaeve <strong>and</strong> Bessière, 2011b].<br />

7.1 What is strategy? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118<br />

7.2 Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119<br />

7.3 Perception <strong>and</strong> interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120<br />

7.4 Replays labeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122<br />

7.5 Build tree prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129<br />

7.6 Openings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137<br />

7.7 Army composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146<br />

7.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156<br />

117

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!