Information Technology Reference
In-Depth Information
Moreover, the fluent graph, and thus the distance function, reflect what could be
called “strategic positioning”: states with pawns on the side of the board are worth less
than those with pawns in the center. This is due to the fact that a pawn in the center may
reach more of the 8 possible destinations than a pawn on the side.
6
Evaluation
We first measure the generality and time requirements of our approach for 201 games
available at http://ggpserver.general-game-playing.de/.
Figure 5 shows the number of games grouped by the minimum time limit (in seconds)
required for successful fluent graph construction.
Fig. 5.
Number of Games grouped by the minimum time limit required for construction of the
fluent graph
We can see that the approach is able to find distance features in all but 5 games.
Construction is typically fast and takes a few seconds for the majority of games. There
are, however, 33 games that require at least 5 seconds for fluent graph construction
and 22 of these even more than 50 seconds. Our interpretation of the results is that the
approach is general, although time constraints may pose a problem in some cases and
should be addressed, e.g., using time-outs.
For evaluation of the playing strength we implemented our distance function and
equipped the agent system Fluxplayer [9] with it. We then set up this version of Flux-
player (“flux distance”) against its version without the new distance function
(“flux basic”). We used the version of Fluxplayer that came in 4th in the 2010 champi-
onship. Since flux basic is already endowed with a distance heuristic, the evaluation is
comparable to a competition setting of two competing heuristics using distance features.
We chose 19 games for comparison in which we conducted 100 matches on average.
Figure 6 shows the results.
