YAFiMM - Yet Another Financial Market Model

The program yfimm reads the description of the market and the agents from a configuration file. The name of this file can be provided at the command line with the option -F. If no file name is provided, yafimm reads the market and agents description from the standard input.

The structure of the configuration file is a sequence of lines. Each line constitutes a record whose field are separated by blank (spaces or tabs) characters. Each record describe one "building block" of the model: the general properties of the market or the characteristic of a group of agents. The generic structure of a line can be represented as

typeofrecord first_arg second_arg ...

The first field of a line specifies the type of record and can assume the following values

market

the line describe a market

trendfollower

the line describe a group of "trend following" agents

fundamental

the line describe a group of "fundamental" agents

sophfund

the line describe a group of sophisticated "fundamental" agents

noise

the line describe a group of "noise trading" agents

(Only the first three characters are relevant to identify the type of a record). Line beginning with a "#" are considered comments and are skipped.

If the record describes a group of agents, the second field has a special meaning: it is the number of agents belonging to this group. The rest of the line is formed by couples of the form key=value. The possible values for the key depend on the type of record and are listed below, together with their default value and constraints

The value can be substituted by a comma separated couple min_{value,maxvalue}. In this case each agent of the group will be assigned a random value drawn from an uniform distribution with support (min_{value,maxvalue}]. Notice that shortA=1 implies shortB=1.

This is the situation: the market is a batch auction; the riskless return is $.001$ and the dividends per share paid at each market round are $.001$. On the market there are \(100\) agents acting as trend followers. They use an EWMA rule with parameter $.97$ for future return forecasting and their risk aversion is randomly chosen between \(100\) and \(200\). During the trading, they send market orders (instead of limit order) to the auctioneer \(50\%\) of the times. Operating on the market there are also \(50\) fundamentalists, whose forecasting rule for return supposes a "market efficiency" of $.5$ and an EWMA coefficient of $.94$. Their risk aversion is equal to 10. Moreover, they receive random shocks, uniformly distributed in \([-2\%,2\%]\), to their future price prediction and they always partecipate the market via limit orders. Both types of agents are initialized with forecasts extracted from a random distribution of mean zero and variance \(0.01\). Fundamentalists can attain short position in bond (i.e. they can borrow money).

This situation described above is represented with the following lines

market type=1 d=.001 r=.001 noise=.01
trend 100 lambda=0.97 beta=100,200 eta=.5
fund 50 theta=.5 lambda=0.94 beta=10 Dy=.02 shortB=1

Try the three .ini files provided with the package as an example.

yafimm is written in C++. It requires the GNU Scientific Library, which should be available through the package management software of any Linux distribution. Remember to install the development version of the libraries before trying to compile yafimm.

Technical information for advanced users (regular users can skip it): The standard packages assume the use of the CBLAS interface provided by the GSL library itself. If you want to use a different library, you should intall via the source code (tar.gz file) and change the LOADLIBES parameter in Makefile to reflect your preferences.

Look at the short help provided by the programs

yafimm -h
deterministic -h
aggregate -h

There is a still-work-in-progress manual in the directory doc. Most importantly, read the papers listed in the References below, where the model and its features are (hopefully) clearly explained.