Re: What is the best way to visualize what happened in a game after it's over?
Towards the coming demo competition:
Does anyone know what is the currently best way to visualize what happened in a game after it's over?
Good question, Daniel -
The existing logtool will let you pull out your choices of events from a state log. From that, you can generate various kinds of single-game plots. You can pretty easily write scripts around it that will do various types of cross-game analyses. The logtool will also let you pull out data and dump it into a database, from which you can do whatever cross-game analyses might be of interest.
I started work on an interface between the logtool and the game visualizer, but it has not been at the top of my priority list for some time, partly because it seemed somewhat inconvenient to have to use a browser to see the data. I do not anticipate having time to revisit this until sometime this summer, probably. But I think it should not be too difficult, and if someone wants to try it, I would be happy to collaborate on it.
I welcome other ideas. It would be nice, for example, to have a tool that could retrieve and unpack game logs on the fly, to save you the pain of having to download them. That might make the web-based approach much more attractive. Perhaps one could build the logtool into a servlet.
I'm sorry if this does not answer your question. It's an important question, and it needs an answer. But it also needs resources.
Better late than never: For our AAAI paper I created a database model and related procedures to populate it from state log files. While that in itself is no visualization, it's easy to create graphs for single games or even across games from that representation using the plotting tool of your choice (I've successfully used matplotlib/Python and ggplot2/R, for example)
We have a small publicly hosted installation of that database, and I'm in the process of loading the April pilot results into it. In addition, I'm interested in working with beta users to get the database model and related documentation to the point where other teams can host their own private installations.
If you're interested you can contact me at firstname.lastname@example.org and we'll figure out the best way to get you started.