FMS is (c) 2008-2012 Jean-Charles Bagneris. See the LICENSE file in the distribution bundles for redistribution information and usual disclaimer.
FMS is an agent-based financial market simulator. The intended audience is financial markets researchers and experimentators, looking to simulate various agents behaviours on different types of markets through the resulting transactions on a fictitious asset. Agents, markets and the environment (the “world”) are Python classes, derived from abstract ones provided with FMS.
As the resulting output (the transactions) is in comma separated values format, it is easy to use it as an input for whichever processing needed (produce graphics, import in spreadsheet, crunch in various statistical procedures, ...)
FMS is a command line application, configured through very simple flat files. As you may write your own agents, markets, engines and world classes, it is as customizable as it could be. But you do not have to be a programmer to use it: FMS is provided with a comprehensive set of classes ready to use, and there is more to come.
If you program your own classes, remember these may be of interest for others: feel free to drop me an email and to contribute (see How could I contribute ? below).
FMS is a simulation tool intended for research only. Thus, FMS is NOT (and will not be in the foreseeable future):
|||Although FMS intention is not primarily pedagogical, it might be useful in classroom environment with PhD students, for an example.|
FMS was primarily developed for my own research projects. The idea came from many other agent based simulation programs, but the design was especially inspired by Julien Derveeuw thesis (in French) : Derveeuw J., Simulation multi-agents de marchés financiers, Université des Sciences et Technologies de Lille, 2008 (see http://cisco.univ-lille1.fr/papers/ for more information).
FMS is slightly different than Derveeuw’s platform in some ways, mainly because FMS can simulate a multi-stages market, e.g. with a pre-opening period where orders accumulate, then a fixing, then a continuous order driven market, starting with the remaining orders after fixing. In addition, FMS is fully open-source (as far as I know, Derveeuw’s platform is freely usable, but sources are not available) and written in Python, which (in my opinion) is easier to learn than Java for researcher whose primary concern is finance, not computer programming.
To use FMS, you first need to download and install it on your system (obviously). See the installation documentation and do not forget to run the tests once you are done.
Then, you should describe an experiment for FMS to run. Experiments are described in config files in the YAML format, which is hopefully rather easy to read and write.
Once you have written an experiment configuration file, you simply run fms on it. FMS will output transaction data on the console or in a comma separated value file, depending on the configuration.
If you read all this, then you certainly have a good reason to use FMS. If the world, engines, markets and agent classes included in FMS do not meet your needs, then you may either write yours, or even (politely) require us to write those for you. Of course, your problem has to be interesting enough for us to do this, and the resulting classes would be part of FMS next release. By the way, if you write yourself an interesting class for FMS, please submit it for inclusion (you would of course be credited for your work). Actually, FMS already includes some contributed classes, see the contrib directory (and the README file in it).
Report bugs, write new classes, translate documentation, write documentation and additional examples, request new features, use FMS and let people know you use it. Think of other ways to contribute. Thank you :)
git clone git://github.com/jcbagneris/fms.git