I am looking for Tools/Software/APIs that will allow me to automatically calculate mixed-strategy Nash Equilibrium for repeated games. I am not looking for trivial solutions to 2x2 games.
My work involves the Leader-Follower model where the strategy-space may increase at every stage of the game.
Hi Patrik,
I really like Gambit, I discovered these libraries when I was researching at Stanford and they performed well for me. As you will see, they are an open-source collection of tools for doing computation in game theory. I can't tell you if they are for heavy-duty stuff.. but I am pretty sure they will work for medium-large size games.
The web page where you can find the libraries:
http://gambit.sourceforge.net/
The citation:
McKelvey, Richard D., McLennan, Andrew M., and Turocy, Theodore L. (2013). Gambit: Software Tools for Game Theory, Version 13.1.1. http://www.gambit-project.org.
Best,
Yago
Hello Patrik,
I have not used specialized tools for Game Theory but since they are mainly approached through Mathematical Programming (MP) I would have used something like AMPL (there is a free version); additionally, I expect MATLAB 2014 to have incorporated some relative (MP) features.
I also recommend Gambit. The code is mature and has been around for a long time (and is well-maintained), it has a GUI that you can use to hand-craft games yourself, and it has a standard file formats to read (e.g. automatically generated) games. It also has a suite of modern algorithms to choose from. Also, in my experience the people on the gambit-users mailing list were quite responsive with questions.
Gambit is great, I used it for a course I was teaching as well as my own research.
http://gambit.sourceforge.net/
I too have had good luck with Gambit. See the article in the 2011 volume of the Annals of Regional Science by Eaton and Tweedle for our application. The software is applicable to problems in which there is a finite set of pure strategies.
Gambit is cool and very well proven. I once implemented the Nash Equilibrium algorithm myself just to cross-validate the results I got from Gambit. Yes, Gambit is very accurate. It is also very useful in the learning process when you just need to see how games are constructed either in normal form or extensive games. I also found out that it supports a Python and C++ languages so you could easily integrate the games in your application just incase solving the game is not the primary goal.
There is a new tool called: Game Theory Explorer β
you can find it here:
http://gte.csc.liv.ac.uk/gte/builder/
Gambit is a very nice suggestion. I was not aware about it till it is mentioned here. I test it successfully. Thank you.
Dear Pratik,
See the following web page:
http://gambit.sourceforge.net/
It is free and open-source software for game theory applications. Try it!
Regards,
Morteza Shabanzadeh
http://mortezash.blogfa.com/
- Badges
- Science topic