trader-desk/doc/trader-desk.texinfo

\input texinfo   @c -*-texinfo-*-
@c %**start of header
@setfilename trader-desk.info
@settitle Trader Desk
@c %**end of header
@copying
Trader Desk

Copyright @copyright{} 2017  The @emph{trader-desk} developers

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.  A
copy of the license is included in the section entitled ``GNU Free
Documentation License''.

A copy of the license is also available from the Free Software
Foundation Web site at @url{http://www.gnu.org/licenses/fdl.html}.

@end quotation
@end copying

@titlepage
@title Trader Desk 0.1
@subtitle Trading robot developer environment
@author Dale Mellor
@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@ifnottex
@node Top, Foreword, (dir), (dir)
@top Trader Desk 0.1

@insertcopying
@end ifnottex

@c Generate the nodes for this menu with `C-c C-u C-m'.
@menu
* Foreword::
* Introduction::
* Installation guide::
* User Guide::                  Invoking trader-desk.
* Algorithms::
* Developer::
* Time Line::
* Copying This Manual::
* Index::
@end menu

@c Update all node entries with `C-c C-u C-n'.
@c Insert new nodes with `C-c C-c n'.
@node Foreword, User Guide, Top, Top
@chapter Foreword

@section Set up to Fail

@subsection The Evil That Men Do

There are many things wrong with the world.  The worst three, in our
opinion from our haughty point of view in the West, are third-world food
shortages, the buy-to-let property industry, and the global stock
exchanges.  The first problem is difficult, but the latter involve able
people sitting on their back-sides while actively screwing other people,
able or not, out of their own money, contributing nothing to the useful
output of any economy.  At least the latter sin offers the prospect for
little people (like us) to shamefully bury their scruples and suckle at
the nipple of big business, and take just a tiny amount of glut for
themselves.  Or, more probably, not.

For we are the worst traders on Earth.  It all started when times were
roaring, before the financial crisis and the fall of @emph{big} banks:
one day we bought shares for the first time, sold them later that same
day, and made a hundred pounds in profit there and then.  Could life get
simpler?  Unfortunately, such times have never returned, and over the
years, we have come to realize that the only way to make money is to
invest long-term in a solid (i.e. massive) company, and you will
hopefully do better than the meagre interest on a bank savings account.
Or maybe not.

@cindex financial crisis
Unfortunately, after being burned badly by the financial crisis and lost
a lot of money (for us), we've been unable to resist the urge to try to
make a quick buck here and there, and, though we no longer dabble in the
folly of amateur day-trading, still play at trying to out-fox the market
and make a quick profit: buying one day in the hope that the price will
go up in the next few days to be able to pull a quick profit; it rarely
does.

Being not very good traders, we've never bothered doing any actual
research in the companies we've bought in to, but go on historical
market data alone to decide where our next gamble will be.  Thus, in
those years of trading we have found the need to assimilate market data
in a local database so that we can explore at will, quickly.  We also
found the need to visualize the entire market, as well as explore
individual positions interactively and with great flexibility of
presentation.  Finally, we have the desire to build us an artificial
intelligence to inform our trading decisions, based on the collected
historical market data.

Applications and widgets abound which display stock market data every
which way you can, and each way seems to show the data in a different
light (a stock looks good to buy one minute, but hopeless when viewed
another way in a different application).  This application aims to allow
for a more interactive exploration of market data, a maximal overview of
the whole market, and to provide a platform for artificial intelligence
algorithms to be developed to identify good buys and sells.

@subsection Trading (don't)!

There's a simple reason why you can't out-fox the markets: if a perfect
algorithm existed everyone would use it and nobody would win.  As it is,
the banks and large agencies have far more resources than you to develop
good (very, very, good) algorithms, to the extent that if they ever find
out what algorithm you are using they will use the knowledge against
you: work out what your next move will likely be, and move to capitalize
on your mis-fortune.

The markets are like a very complex game of chess.  It is warfare.
Dog-eat-dog warfare.  It is your own real, hard-won, money that you will
lose when you are defeated in battle.

For this reason there is no point in us telling you the secret to our
(lack of) success.  Nor is there any point in you taking anyone else's
advice.  In fact, the large agencies are apt to give out lots of advice
to investors knowing how the advice itself will affect the markets and
moving to capitalize themselves.  Never bother following @emph{The
Motley Fool}.  Wish we didn't have to mention them.  For this reason
@code{trader-desk} does not come with any trading strategies built in.
That would just be us shooting ourselves in the foot.

The only chance you have of success is by developing a unique strategy
of your own and keeping it to yourself.  The ultimate aim of
@code{trader-desk} is to help you to develop that strategy, and to
implement it in code.

@subsection Caveat Emptor

What you do in the stock markets is entirely up to you.  We've already
tried to dissuade you from trying to play the game and get rich quick.
If you do, and you use this application, that's your choice.  But your
ultimate actions are nothing to do with us.  We hereby divulge ourselves
of all responsibility for your future well-being through use of this
application (or otherwise, actually).

@section The @code{trader-desk} Application

@cindex trader-desk
This introduction to the @code{trader-desk} application has been
deliberately meaty.  We've tried our best to persuade you that using
this application is not a good idea, for the reasons given above.
However, if you've come this far and we haven't frightened you away,
welcome.  We'll lighten the tone from now on, and get to thinking
positive!

@contents

@node Introduction, Installation guide, Foreword, Top
@chapter Introduction

This manual pertains to pre-release version 0.1 of the
@code{trader-desk} application.  Despite the lowness of this number
and the fact that it is the very first release, do not underestimate
the amount of effort which has already been invested in this project!

@cindex alpha software
First, understand that this is the very first release of untested
alpha-grade software.  It has some arbitrary limitations (availability
of market data, UK-centric outlook, lack of customizability), and some
idiosyncrasies that will be mildly annoying.  Read the @ref{Time Line}
chapter for further information on the current state and direction of
the code base.

@cindex bug reporting
To report bugs or make contact with the authors please go to
@url{http://rdmp.org/trader-desk/contact}.

We take such pride in our work that hopefully there aren't any hard bugs
as such: ones which cause the program to shut down or become terminally
unresponsive, or corrupt the local database.  But no guarantees.

@node Installation guide, User Guide, Foreword, Top
@chapter Installation Guide

The software has been successfully installed on Debian Jessie (stable),
Stretch (testing), Ubuntu 16.04 LTS, Ubuntu 17.04 (all Debian derivatives)
and Fedora 25 Workstation (a Red Hat derivative).  We've also tried to
use Scientific Linux 7, but that is so far behind the curve that
@code{trader-desk} cannot be compiled here.

The build is actually a very standard @emph{autotools} procedure:
@code{./configure --with-database-password=xxx; make; make install;},
hence it should be possible to build on most POSIXish operating systems.
Instructions below are for the select specific Gnu/Linux distributions
mentioned above, and may need to be adapted for other systems.

@menu
* System Preparation::
* Database Preparation::
* Trader-Desk Installation::
* Issues::
@end menu

@node System Preparation,  , Installation guide, Installation guide
@section System Preparation

At the command line (you will need @code{sudo} privilege on the system,
and may be prompted -- just once -- for your login password).

@enumerate

@item  Fedora 25 Workstation
@example
sudo yum install mariadb mariadb-server mariadb-devel gettext-devel \
                 patch gcc-c++ libcurl-devel gtkmm24(-devel?)
sudo systemctl start mariadb
@end example

@item  Debian stable (Jessie)
@example
sudo apt install build-essential mariadb-client mariadb-server \
                 libmariadb-client-lgpl-dev gettext pkg-config curl \
                 gtkmm-2.4 libcurl4-openssl-dev
@end example


@item  Debian testing (Stretch)
@example
sudo apt install build-essential mariadb-client mariadb-server \
                 libmariadbclient-dev gettext pkg-config curl
@end example


@item  Ubuntu 16.04
@example
sudo apt install autoconf autopoint libtool mariadb-client \
                 mariadb-server libmariadb-client-lgpl-dev gtkmm-2.4 \
                 libcurl4-openssl-dev curl
@end example


@item  Ubuntu 17.04
@example
sudo apt install autoconf autopoint libtool mariadb-client \
                 mariadb-server libmariadbclient-dev gtkmm-2.4 \
                 libcurl4-openssl-dev curl
@end example


@end enumerate

@node Database Preparation,  , System Preparation, Installation guide
@section Database Preparation

On all systems, at the command-line:

@example
sudo mysql -u root
>>  create user trader_desk@@localhost;
>>  grant all privileges on *.* to trader_desk@@localhost;
>>  exit;
mysql -u trader_desk
>>  create database companies;
>>  exit;
@end example

@node Trader-Desk Installation,  , Database Preparation, Installation guide
@section Trader-Desk Installation

On all systems, at the command-line:

@example
curl https://rdmp.org:9443/trader-desk/trader-desk-0.1.tar.gz | tar xzf -
cd trader-desk-0.1
cp trader-desk/example-config.h trader-desk/config.h

[ On Fedora only: edit trader-desk/config.h and change the
                  DATABASE_SOCKET to /var/lib/mysql/mysql.sock. ]

./configure --with-database-password= --without-boost
make
sudo make install
@end example

@subsection Variations

The above instructions are deliberately the minimum you need to get
going.  However you may need to vary these for your installation:

@enumerate
@item

Use a @code{--prefix} option on the @code{configure} line to specify
where to install the @code{trader-desk} application.  You would need to
do this if you don't have @code{sudo} access (the default is to install
under @code{/usr/local}).

@item

Set a password on the @code{trader_desk} database account.  Also, if the
@code{root} account has a password set, this will need adding to the
@code{mysql -u root} line.

@item

Make customizations of the @code{trader-desk/config.h} file.  For
example, extend the date range for which market prices are initially
obtained.

@end enumerate


@node Issues,  , Trader-Desk Installation, Installation guide
@section Issues

If you are using Ubuntu, the `Update' menu does not become inactive when
a single company is in the main display; using the items on this menu
when a single chart is displayed in the main window WILL cause the
program to crash.

@node User Guide, Algorithms, Installation guide, Top
@chapter User Guide

@menu
* Getting Started::             Running the program
@end menu


@node Getting Started,  , User Guide, User Guide
@section Getting Started

@cindex Running the program
@example 
trader-desk [--version | --help]
@end example

@cindex version
Unless you are simply looking to find out the version of the installed
program (using the --version option), simply running
@code{trader-desk} will start the program and produce a new window on
the screen.

As soon as the program starts you will be presented with a number of
markets in which you may take an interest.  Double-click on one of
these markets.  It will take some time to ingest a few years' data for
the entire market; be patient.

Now, maximize the window, enjoy your overview of recent market
movements, click on one of the components (companies), and move the
mouse around...


@menu
* Main screen::
* Detailed analysis::
* Preferences::
* Menus::
@end menu


@node Main screen, Detailed analysis, Getting Started, Getting Started
@section The Main Screen

This screen shows the last 50 days' prices for all components in a
market, so that you can see at a glance which stocks are overall up and
which are overall down, and, more to the point, which are turning.

The menu is fully serviceable on this screen: data from other markets
can be obtained from the `File' menu, the components of a different
market can be shown from the `Display' menu, and the most up to date
data can be obtained for the current market from the `Market' menu.

Finally, to analyze a company's stocks in more detail, simply click on
the appropriate chart.

@image{grid,,,Screen with thumb-nail chart for every company,png}


@node Detailed analysis, Preferences, Main screen, Getting Started
@section The Detailed Analysis

This screen shows the price data for a single company, and the user can
use the mouse to change the nature of the display, show precise stock
values at arbitrary times, and indicate a rectangular region across
which differential values are shown.

Note that the window as you see it may differ slightly from that shown
below due to cosmetic thematic variations in the appearance of
workstation windowing systems.

@image{detailed-analysis,,,View of window with numbered place markers,png}

@enumerate

@item
Click on these arrow buttons to cycle through the companies in the
current market in alphabetical order.  Once the last company is reached
the cycle will wrap around to the first one, and similarly if cycling
backwards from the first item.

@item
This panel serves both to indicate which company's data are being
displayed in the chart, and to allow the user to select an arbitrary
company in the current market: start typing here and a drop-down box
will appear with possible completions of company names.  Once selected,
press the ENTER key to load that company's data.

@item
This drop-down box will show all open positions you have in the present
market (if any).  Selecting one will: load the data for that company,
move the shares scale to indicate the number of shares you are holding,
and will indicate the position in (time, value) space at which the
position was opened, with a green vertical bar and tide line.

@item
When a new chart is loaded this panel shows the latest price per share
of the commodity.  It can be edited, and then on pressing the enter key
the new price will appear at the current time on the graph, allowing the
user to manually input an up-to-date price.  He would especially want to
do this prior to clicking the ‘BUY’ button as this price will then be
entered as the purchase price of the newly opened position.

@item
This button will show either `BUY' or `SELL'.  If there is an open
position on the stock being displayed, pressing the button will update
the position as being closed at that time, at the price per share shown
at (4).  Otherwise, pressing the button will create a record of an open
position, using the unit price at (4) and the number of shares at (9).

@item
This axis shows the real value of the shares, taking into account the
number of shares indicated at (9); it will change dynamically as that
slider is moved.

@item
Dragging the mouse (holding the primary button down while moving) over a
chart will cause a rectangular region to be highlighted, with
information around the edge to indicate the increments in the various
quantities that the lengths of the edges represent.  To remove this
informative device, click anywhere on the chart.

@item
When the mouse cursor is over the chart, cross-hairs will provide
precise information about times and stock values at that position.  Note
that two numbers are generally displayed: to the right is the unit price
of the stock, in pence/cents, and to the left is the full value of a
position taking into account the number of shares at (9), in
pounds/dollars/euro.  Also note the precise time displayed just above
the bottom axis of the chart.

@item
Here you will see an indicated number of shares, used to work out the
full value of any (potential) positions that might be opened, and to
work out the profit line at (14).  In the case that there is a position
open on this stock, the slider itself will not be present but will
represent the number of shares in this position (the slider can be
restored by selecting `NONE' at (3)).

When a new position is to be recorded, this slider should be used to set
the precise number of shares in the purchase, before pressing the buy
button at (5).

@item
This slider alters the temporal span of the chart, allowing for the
selection of a long-term or short-term view of the stock's value's
evolution.

@item
This slider selects the size of the window used to calculate the mean
data line (red line on the chart): a larger window means the line will
be smoother.

The size of the window is indicated at (15).

@item
This slider controls the width of the envelope (yellow area), so that
the user can determine a useful measure of the variability of the data
and the degree of extremity of the current position.

@item
This is a very dynamic scale showing the calendar time at which prices
were obtained.  If weeks are being indicated, the dates shown will be
those of the Mondays.

@item
The dark gray tide is the only one whose meaning is not obvious.  It
represents the profit line for a position in this stock.  Taking into
account the cost of trading (set in the preferences) this is the point
above the current value at which those trading costs become offset by
the increased value of the stock (taking the number of shares, set at
(9), into account).  If there is an open position, the line will
represent the real profit line for that position.  Otherwise it
represents the profit point if the stock were purchased at the latest
price.

@item
This vertical line represents the size of the moving-average window used
to produce the smoothed time-series show in red on the chart.  The
distance to the right-hand edge of the chart is half the window width:
practically it means that the data-average line to the right of this
vertical line is computed over a shorter window than the rest of the
graph, making the line more erratic there, increasing so as the
extremity of the chart is reached.


@end enumerate


@node Preferences, Menus, Detailed analysis, Getting Started
@section The Preferences Dialog

The preferences dialog allows the user to set parameters which affect
the whole of the @code{trader-desk} application.  Note that there is no
`SAVE' button here; pressing the enter key or dismissing the box will
make the settings entered take effect immediately.

@image{preferences,,,The preferences dialog box,png}

The items are

@enumerate

@item Fixed cost of trading

This should represent the fixed cost of buying @emph{and} selling shares
(respectively, opening and closing a position).  It is expressed in
pounds, but you may need to take this as dollars or whatever your major
currency unit is.

@item Proportional cost of trading

This should represent the proportional cost (if any) of buying
@emph{and} selling shares, expressed as a percentage.  It is usually
used to account for stamp duty, but may also apply to agent or market
fees.

@end enumerate

@node Menus,  , Preferences, Getting Started
@section The Menus

@itemize
@item
File -> Ingest new market

This will bring up a dialog box allowing for the selection of a market
whose data are not currently being tracked by this application.  Once
selected, the progress of downloading historical data for that market
will be indicated in the displays, but expect this to take some not
inconsiderable amount of time.

@item
File -> Preferences

Select this item to bring up the preferences dialog, allowing for the
specification of parameters which affect the overall operation of this
application.

@item
File -> Quit

Select this to exit the program.

@item
Display

Selecting this menu item will produce a list of all markets for which
historical data are currently being held.  Select one of those items to
see a market overview: thumbnail charts for all components in that
market.

@item
Market -> Update latest data

Selecting this will cause (after a short delay) the download of the
latest prices for all the components in the current market.  Note that
this option will be inactive if the display is not showing a market
overview.

@item
Market -> Update close data

This action will cause a fetch of all market closing prices for all
components which are not currently in the database.  This is a slow
process, but progress will be indicated through the real-time update of
the market overview, and a scroll bar in a small pop-up dialog box.

@item
Help -> About

This option will produce an informational dialog box, displaying the
current state and credibility of this program.

@end itemize


@node Algorithms, Developer, User Guide, Top
@chapter Hints for Algorithm Developers

As previously alluded to, we don't supply any trading algorithms with
this product.  Here we give some general hints to the top-level approach
to developing algorithms.

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

@section Comparing algorithms

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

The basic idea is that you think of two possible approaches to making
trading decisions based on all the current data for which you have
historical records.  Run the two algorithms over the history of data you
have (i.e., pretend that you have been trading stocks and shares for all
that time), and compare the amount of money that each algorithm makes or
loses.  Choose the most favourable one.

Now take a close look at how the best algorithm performed, given the
data, and try to think of a way to improve the algorithm.  Now iterate
back to the paragraph above and compare the new algorithm with the
previous best.  Either accept the new algorithm, and try to improve
that, or reject it and go back to the drawing board.

@section Iterate, find the best algorithm

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

The next level of sophistication is to let the computer perform the hunt
for the best algorithm.  This requires that the algorithm uses one or
more parameters as part of its definition.  For example, you might
decide to buy the stock for which the average difference between the
previous @code{n} prices is biggest, out of all the available
companies on the market.  This example has the single parameter,
@code{n}, which can take on any of the values 2, 3, 4, ....  Choose a
starting value, such as @code{n}=5.  Now get the computer to compare
the returns with @code{n}=4 and @code{n}=6, and see if either are
better than @code{n}=5.  If not, choose @code{n}=5 as the best
algorithm.  If @code{n}=4 was best, have the computer next try
@code{n}=3, or if @code{n}=6 was best, try @code{n}=7.
Etcetera.

If your algorithm has more than one parameter, it may become expensive
to try all combinations of parameter variations about the vicinity of
your starting point.  In this case you will have to try a smaller number
of nearby points chosen at random, using a computer's pseudo-random
number generation facilities.

@section Using data

The more information you have available to guide your dealing choices
the better.  This program provides you with a daily history of closing
prices, for an entire market.  However, if you can also draw on other
data in real time, you should factor those into your decision algorithm.
Unfortunately, you will always be at a disadvantage compared to
professional traders with regards to news sources, as they pay stupid
money to get the news delivered to them first.  Or perhaps you can think
of a way to make old news work against the big boys... a bit of reverse
psychology, perhaps?

@menu
* Bayes::
@end menu

@node Bayes,  , Algorithms, Algorithms
@section The Bayesian Programme

@subsection Bayes' Theorem

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

The modern way to think in science is to realize that, going in to any
experiment, we already have a good idea of what the outcome is likely to
be, and that the effect of performing some investigation is to provide
data which allow us to update, or refine, our initial idea based on the
results.  We also take the view that we never know the true outcome of
our experiments.

@cindex Bayes' Theorem
It starts with the simple idea that the probability of two events
happening is the probability that the first event happens multiplied by
the probability of the second event happening knowing that the first
event already happened: P(A and B) = P(A)P(B|A).  But the argument could
equally well have been made the other way around, and so P(A)P(B|A) =
P(B)P(A|B).  Now re-arrange this to give P(B|A) = P(B)P(A|B)/P(A).  Now
consider that the event A is the outcome of some experiment, and the
event B is the probability that a model parameter has some specific
value.  Then the above equation states that the parameter value given
the experimental result is equal to the parameter value probability
before the experiment multiplied by the probability that the
experimental result came from a model with the original parameter
probability and divided by the probability of the experimental result.
In other words, it is a recipe for updating original knowledge of a
parameter probability with the result of an experiment to measure that
parameter.

[Apply to market modelling.]

@subsection The Metropolis Algorithm

@node Developer, Time Line, Algorithms, Top
@chapter Notes for Software Developers

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

@menu
* GIT::
* Development Style::
* Analytic Tools::
* Trading Robots::
@end menu

@c @node GIT, Development Style, Developer, Developer
@c @section Sources from GIT

@c @cindex git
@c @cindex gtkmm
@c @cindex gcc
@c @cindex bootstrap
@c The procedure is essentially the same as for the user instructions
@c above--including the setting up of a suitable database--, except for
@c the extra boot-strapping step, which will require a suite of developer
@c build facilites (autoconf @emph{et al.}) to be available on your system.  The
@c development has been done on Ubuntu-1610, Debian stable and Guix,
@c though the latter two require some choice package selections to be
@c made (GTKmm and GCC need to be 2.x and 4.9.x, respectively).

@c @example
@c git clone .../trader-desk
@c ./bootstrap.sh;
@c [Edit trader-desk/config.h]
@c ./configure --prefix=... --with-database-password= --without-boost
@c make
@c make install
@c @end example


@node Development Style, Analytic Tools, Developer, Developer
@section Development philosophy

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

@subsection Open source

While we have nothing against the @emph{publish early, publish often}
philosophy, it is not for us at this time.  While the project is in the
early stages of development we are taking things carefully, slowly and
methodically and following our own @ref{Time Line}, publishing only at
planned islands of stability.

We believe that this is most conducive to creating the project we want,
and allows meaningful testing to take place without wasting too much of
people's time.

The intention is that the project will ultimately be owned, and a part
of, the GNU project.  The intention is also that the hand-over will not
occur before completion of the initial development and establishment of
the code as being of production quality.  In the meantime, we keep
ownership to ourselves, but all releases will be made under the
GPLv3+.

@subsection Programming

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

@cindex guile
We are massive fans of Guile and the Guile-as-extension-language
paradigm, and given that the ultimate aim of this project is to provide
a base for the development and deployment of artificially intelligent
robots for making trading decisions, it would seem natural that most of
the project would be written in Guile Scheme with only some essentially
low-level core parts written in C.

However, it is the intention that this application allows robots to run
at least an order of magnitude faster than real-time (so that they can
assimilate the entire history of the market and make timely
recommendations for activity), and allows users to explore the market
data and results of analysis with buttery-smooth interaction so that the
mechanics of the software and data management do not interfere with the
exploration and research of ideas.

@cindex c++
Thus the application is being developed purely in C++ (version 14 no
less, and fully version 17 by the time the project reaches production
release).  This language provides the performance and facilities we
desire, provides the mechanisms to imbue innate structure to the
software, and provides the means to specify clean, verifiable,
interfaces for external modules.

As to the programming style and project presentation we stick to the GNU
coding standards and maintainer guidelines as far as practically
possible (allowing for the fact that the standards only truly apply to
software written in C).  Build management is achieved through Autotools,
and package management is offered in the form of GUIX scripts.

@subsubsection External dependencies

We aim to keep external dependencies to a minimum, but these are the
ones we currently use which are readily available to most GNU
systems.

@cindex gettext
@itemize @bullet
@item 
@code{gettext}: for linguistic localization of the package;

@cindex mariadb
@item 
A running @code{MariaDB} database and client library: to store locally
and to query historical market data;

@cindex gtkmm
@item 
@code{GTKmm}: the graphical widget toolkit on which the GUI is built (we
are currently stuck at 2.4; the move to version 3 will be expedited as
soon as we get the chance);

@cindex curl
@cindex libcurl
@item
@code{libcurl} on some TLS implementation: to access services on the
Internet.

@end itemize

@cindex curlpp
@cindex quote library
@cindex bootstrap
Additionally, we use @code{CURLpp} (to access libcurl through a C++
interface) and the @code{quote} library (for access to the Yahoo!
Finance API) which are seldom so readily available; these are packaged
into distribution tarballs with the project sources, @c and a GIT bootstrap
@c script is provided to pull the sources into a development project tree
wherein they will be built and integrated through the Autotools
mechanisms.

@subsection Data sources
    
At the time of writing the only freely accessible and reliable source of
market price data, both current and historical, is Yahoo!  Finance.  It
is however absolutely not the intention of this package to nail itself
to Yahoo!, and even though there are currently no alternatives the
interface to the price server is to be fully abstracted to allow for new
possibilities in the future.

Unfortunately, Yahoo! does not provide a service which tells the
components (companies) which make up a market, and their symbols are
incompatible with those from other services (there is no standard,
strangely enough).  Our solution is to provide a service at
@url{https://rdmp.org/trader-desk} to provide the information, and which
we are currently maintaining, by hand, ourselves (although the process
will become more automated as we gain experience).  Again, it is the
intention that this project's interface to the RDMP service will be
fully abstracted, so that other back-ends can conceivably be developed
in the future.

@node Analytic Tools, Trading Robots, Development Style, Developer
@section Creating Analytic Tools

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

`Analytic tools' are interactive devices that allow a user to explore
the data of a particular component of the market.  Examples of tools
currently implemented: moving average and standard deviation-based
envelope display, the position analyzer which puts the break-even
tide mark onto all the plots, and the delta-region analyzer which allows
the user to delineate a region on a chart and see the temporal and stock
value changes which that area represents.

These tools must all be derived from the Trader_Desk::Analyzer class.
The collection of tools forms a stack inside the application, and each
time a chart is to be rendered on-screen the stack will be called, in
order, to render the analysis results for the eyes of the user.

Currently there is no way to package tools so that they can be
incorporated at run-time; that will (hopefully) come with the next
pre-release.  To hack a new tool in, read the
@code{trader-desk/analyzer.h}--the interface you need to implement--,
and @code{trader-desk/moving-average-analyzer.h} sources which are
heavily documented and provide a good example which can be used as a
starting point for your own development.  Then edit
@code{trader-desk/analyzer.cc} to cause your new object type to be
realized within the application (this and @code{trader-desk/makefile.am}
are the only original distribution source files you should need to
touch).


@node Trading Robots,  , Analytic Tools, Developer
@section Developing Trading Robots

[This section of the manual is in an early stage of development and is
probably not entirely useful at this time.]

Contrary to the analytic tools described in the previous section,
which work on data of a single company, a trading robot is a tool
which works on the data of an entire market.  As such, it is in a
position to work out optimal trading strategies and make informed
suggestions for the user's trading activities.

At this point in time there actually are no hooks for where such code
would be called out from the program, nor are there any well-defined
interfaces for such beasts.  To get started you should familiarize
yourself with the @code{Trader_Desk::Chart_Grid} object and its member
types: @code{Trader_Desk::Chart_Data} and
@code{Trader_Desk::Time_Series}; these hold the data you have to play
with.  The sky's the limit and the rest is the extent of your
ingenuity.

Proper support for such robots will be along in the next pre-release
(version 0.2).


@node Time Line, Copying This Manual, Developer, Top
@chapter Development Time Line; The State of the Onion

Currently at version 0.1, the package is in alpha-development state and
not at all well tested in the wild (well, it works for me...).

It is anticipated that the following developments will take place before
each release.  Note that there are no implied dates associated with the
releases; you should assume several months will pass between each
release.

@table @samp

@item 0.1

Solid operation on limited data sets, and limited scope for extension
(other than to hack the code).

@item 0.2

@itemize @bullet

@item
Move to GTKmm 3.x and GCC 6+;

@item
Re-factoring of the libraries into functional units;

@item
development of APIs for analytical tools and robot trader extensions;

@item
abstraction of database and market data server back-ends;

@item 
have full set of market data available.

@end itemize

Hopefully the code quality can be upgraded to beta status at this time.

@item 0.3

@itemize
@item
Choice of MySQL/MariaDB or Postgres as database back-end (maybe also
sqlite, ndb, nosql?);

@item
display of historical positions (stock holdings) and API for
retrospective analysis;

@item
indication of weekends and bank holidays with option to contract them on
the charts.

@item
Hopefully the code can be considered fully production quality and made
available to the masses as a full GNU product.

@end itemize

@item 0.4

Abstraction of control interface, with implementations as command-line
utility (usable in scripts) and web-centric service.

@item 1.0

Development complete; we will stay around to maintain the project, but
otherwise it will be left in perpetuity to the FSF as part of the GNU
project.

@end table

Of course, any serious bugs and issues which are reported will also be
dealt with during the release cycle.

As we are currently in alpha development and many changes to libraries
and APIs are in the pipeline, we do not anticipate accommodating major
development efforts from outside.  Minor patches to fix specific,
definite problems in the code will be welcomed, however, as would
proposals for future endeavours.

Contact with the developers about any of the above issues can only be
made through @url{https://rdmp.org/trader-desk/contact}; you may also
make a request here to be added to a mailing list which will inform you
about new releases (this is very low volume).


@node Copying This Manual, Index, Time Line, Top
@appendix Copying This Manual

@c Get fdl.texi from http://www.gnu.org/licenses/fdl.html
@include fdl.texi

@node Index,  , Copying This Manual, Top
@unnumbered Index

@printindex cp

@bye

@c trader-desk.texi ends here