+- Ludii Forum (https://ludii.games/forums)
+-- Forum: Questions (https://ludii.games/forums/forumdisplay.php?fid=13)
+--- Forum: About the Digital Ludeme Project (https://ludii.games/forums/forumdisplay.php?fid=16)
+--- Thread: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht (/showthread.php?tid=852)
Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - Eric Piette - 03-01-2022
--------------------------------------
CALL FOR PARTICIPATION
--------------------------------------
2nd Digital Ludeme Project Workshop: Game AI Applications for Historical Games Research
11-14 April 2022, Maastricht, Netherlands
https://ludii.games/workshop.php
TOPICS
This Digital Ludeme Project (DLP) workshop will gather AI researchers, Archaeologists, Historians, game designers and other games scholars to develop interdisciplinary approaches for a better understanding of traditional games and the development of new AI techniques.
The main topics of this workshop are to discuss overlapping questions, approaches and challenges with regard to Game AI and the History, Archaeology and Anthropology of games, using the Ludii general game system (http://ludii.games/)—and the significant database of evidence for traditional games collected by the Digital Ludeme Project (http://ludeme.eu/)—to investigate issues regarding the transmission, identification, reconstruction and evolution of games.
During this workshop, researchers will present their recent ideas and research interests to motivate discussion and potentially lead to new collaborations. DLP team members will present recent work on reconstructing, analysing, documenting and mapping games, in addition to general Game AI, and will conduct tutorials on using the Ludii system. These presentations are aimed at fostering inter-disciplinary discussions over the four days of the workshop.
Topics of interest include:
- Digital Archaeoludology: Modelling, analysing and reconstructing ancient games.
- Game AI: General Game Playing, search and machine learning.
- Archaeology and the History of Games: Transmission, identification and evolution.
- Game Design: Measuring games for playability and quality.
- Games for Education.
INVITED SPEAKERS
Yngvi Björnsson, Reykjavik, University, Iceland
Tristan Cazenave, LAMSADE, Paris-Dauphine, France
Quentin Cohen-Solal, LAMSADE, Paris-Dauphine, France
Veronique Dasen, University of Fribourg, Switzerland
Thierry Depaulis, independent, France
Eddie Duggan, University of Suffolk, UK
Tiago Hirth, University of Lisbon, Portugal
Fatih Parlak, University of Cappadocia, Turkey
Lisa Rouget, Université de Bretagne Occidentale, France
Jorge Nuno Silva, University of Lisbon, Portugal
Marco Tibaldini, Free University of Bozen-Bolzano, Italy
Ulrich Schaedler, University of Fribourg, Switzerland
Jacob Schmidt-Madsen, University of Copenhagen, Denmark
Hui Wang, LAMSADE, Paris-Dauphine, France
REGISTRATION and PARTICIPATION
The workshop will be a public event with free registration. However, due to a limited number of places available, please contact the organisers by email to take part: ludii.games@gmail.com
The event will also be broadcast on Zoom. The link will be available at: https://ludii.games/workshop.php.
ORGANISERS
Eric Piette, Maastricht University, the Netherlands
Walter Crist, Maastricht University, the Netherlands
Cameron Browne, Maastricht University, the Netherlands
Matthew Stephenson, Maastricht University, the Netherlands
Dennis Soemers, Maastricht University, the Netherlands
Digital Ludeme Project (http://ludeme.eu/)
Funded by the European Research Council (ERC)
Workshop co-funded by the European Cooperation in Science and Technology (COST)
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - Tester001 - 10-30-2022
Hi Eric, I had a look at the Report on the 2nd Digital Ludeme Project Workshop (2022) and found especially the talk of Cameron Browne about 'Game Metrics: Estimating Game Quality' very interesting!
Could something like this be included into the Ludii-player? (like 'Estimate game tree complexity' etc. in the Analysis menue)
I tried to understand it all and simulate it with the Ludii-player, but since one can only set the evaluation time in steps of seconds, this didn't really work out.
The game of my interest is MensaSpiel, because it has a relatively small board but complex gameplay.
With runtime up to some hours available what could I do with the Ludii-player to get something similar like the results in the attachment?
Best Regards, Franz
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - dale walton - 10-31-2022
Interesting read. I think the plot should be 2D: Create a cloud of win vs loss results using random points for 1st player look ahead depth and 2nd player look ahead depth weighted evenly on a logarithmic plot up to a reasonable limit and fit a surface to approximate the shape of equal win - loss. Then look at its shape.
-- Or look directly at the shape of the maximum win density surface.
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - cambolbro - 11-18-2022
Hi Franz,
Sorry for the delay. We've now added Skill Trace to Ludii in the Evaluation menu, so you should be able to run these tests yourself on your own games. This update will be in the downloadable Ludii.jar next time we build a release.
Please find attached an example of the PostScript source files for one of the Skill Trace plots, in case you want to make your own plots.
Regards,
Cameron
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - Michael - 11-18-2022
Amazing! Thanks :)
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - Tester001 - 11-22-2022
Wow!, yes that sounds great! This is definitely something to look forward to.
So when the line in the plot goes up it's a good sign, right?
Will there also be a possibility to set a time-frame for the Skill Trace?
Because for complex games (like chess or go etc, as can be seen in the paper) the 20 seconds my not be enough to include higher BF multipliers.
Also interesting that this test-method works with MCTS (UTC) for games where AlphaBeta ist typically the best search algorithm when given average computing power.
Thx for adding this interesting feature.
Regards, Franz
RE: Game AI Applications for Historical Games Research, 11-14 April 2022, Maastricht - cambolbro - 11-24-2022
Hi Franz,
Yes, it's a good sign when the plot goes up, or even flat can be good if the plateau is well above the zero line. The only thing you *really* don't want is a flat line near the zero mark.
The Skill Trace section of the Evaluation dialog (in the next Ludii release) provides the following options:
1. Levels: Number of skill levels tried.
2. Trials: Number of trials per comparison.
3. Time Limit: In seconds.
The process will stop when the time limit is reached or the desired skill level is reached, whichever comes first.
Yes it's interesting that the method also seems to work for games that are more amenable to Alpha Beta-based approaches (e.g. Chess). I think this is because relative playing strength (between successive skill levels) is more important than absolute strength. Also the continuous "any time" nature of UCT means that we get a meaningful result at the cut-off point for each skill level, rather than having to compute complete plies for Alpha Beta-based approaches. Also, the odd-even effect seen with Alpha Beta can throw off the skill level calculation with wildly zig-zagging plots :)
Regards,
Cameron