beancounter - Stock portfolio performance monitor tool
beancounter [options] command [command_arguments ...]
addindex index args add stock(s) to market index 'indx'
addportfolio sym:nb:fx:type:o:pp:pd ...
add 'nb' stocks of company with symbol 'sym'
that are listed in currency 'fx' to the
portfolio with optional 'type' and 'owner'
info, purchase price 'pp' and date 'pd';
see below for a complete example
allreports combines dayendreport, status and risk
addstock arg ... add stock(s) with symbol arg to the database
advanceement report on unrealized gains from lows
backpopulate arg ... fill with historic data for given stock(s)
dailyjob combines update, dayendreport, status + risk
dayendreport reports p/l changes relative to previous day
delete arg ... delete given stock(s) from database
destroydb delete the BeanCounter database
plreport run an portfolio p/l report rel. to any day
quote arg ... report current data for given stock(s)
retracement report unrealized losses from highs (drawdowns)
risk display a portfolio risk report
status status summary report for portfolio
update update the database with day's data
warranty display the short GNU GPL statement
--help show this help --verbose more verbose operation, debugging --date date report for this date (today) --prevdate date relative to this date (yesterday) --currency fx set home currency --restriction sql impose SQL restriction --extrafx fx1,fx2,... additional currencies to load --forceupdate date force db to store new price info with date --rcfile file use different configuration file --[no]fxupdate enforce/suppress FX update, default is update --dbsystem system use db backend system, default is PostgreSQL --dbname name use db name, default is beancounter
beancounter gathers and analyses stock market data to evaluate portfolio performance. It has several modes of operation. The first main mode is data gathering: both current data (e.g. end-of-day closing prices) and historical price data (to back-populate the database) can be retrieved both automatically and efficiently with subsequent local storage in a relational database system (either PostgreSQL or MySQL though any other system with an ODBC driver could be used). The second main mode is data analysis where the stored data is evaluated to provide performance information. Several canned reports types are already available.
Data is retrieved very efficiently in a single batch query per Yahoo! host from the Yahoo! Finance web sites. Support exists for North America (i.e. US and Canada), Europe (i.e. the Continent as well as Great Britain), several Asian stock markets, Australia and New Zealand. New markets are easy to add by following the (fairly extensively commented) Perl code. Patches are certainly welcome for further markets especially in Asia and Latin America!
beancounter can aggregate the change in value for the entire portfolio over arbitrary time horizons (provided historical data has either been gathered or has been backpopulated). Using the powerful date-parsing routine available to Perl (thanks to the Date::Manip modules), you can simply say 'from six months ago to today' (see below for examples).
beancounter has been written and tested under Linux. It should run under any standard Unix as long as the required Perl modules are installed, as as long as the database backend is found.
beancounter update --forceupdate today
This updates the database: it extends timeseries data (such as
open, low, high, close, volume) with data for the current day,
and overwrites static data (such as capital, price/earnings, ...)
with current data. All stocks held in the database are updated
(unless the --restriction argument instructs otherwise). The
--forceupdate option lets the program corrects incorrect dates
returned from Yahoo! (which happens every now and so often), but
be careful to correct for this on public holidays. Note that
the --restriction argument will be applied to the portfolio table,
whereas the overall selection comes from the stockinfo table.
beancounter addportfolio SUNW:100:USD:joe:401k:85.50:19991117 \
IBM:100:USD:joe:401k:90.25:20000320 \
SPY:50:USD:joe:ira:142.25:20000620
This adds the two stocks Sun and IBM to the 401k portfolio of Joe,
as well as SP500 'Spiders' to his IRA portfolio. The stocks are
also added to the general stock info tables via an implicit call
of the stockinfo command.
beancounter addstock LNUX RHAT COR.TO
This adds these three Linux companies to the database without adding
them to any specific portfolios.
beancounter backpopulate --prevdate '1 year ago' \
--date 'friday 1 week ago' IBM SUNW HWP
This backpopulates the database with historic prices for three
hardware companies. Note how the date specification is very general
thanks to the underlying Date::Manip module.
beancounter plreport --prevdate '1 month ago' --date 'today' \
--restriction "owner='joe'"
This calculates portfolio profits or losses over the last year. It
also imposes the database restriction that only stocks owned by
'joe' are to be included.
beancounter status --restriction "type='401k'"
This shows a portfolio status report with the restriction that only
stocks from the '401k' account are to be included.
beancounter risk --prevdate "6 month ago"
This shows a portfolio risk report. This tries describes the
statistically plausible loss which should be exceeded only 1 out
of 100 times (see below for more details).
beancounter dailyjob --forceupdate today
Run a complete 'job': update the database, show a day-end profit/loss
report, show a portfolio status report and show a riskreport. In the
update mode, override a potentially wrong date supplied by Yahoo!
with the current date.
The following few paragraphs will illustrate the use of beancounter. We will set up two fictional accounts for two brothers Bob and Bill (so that we can illustrate the 'owner' column). The prices below are completely fictitious, as are the portfolios.
We suppose that beancounter is installed and that the 'setup_beancounter' command has been run. We can then create a two-stock (computer hardware) portfolio for Bob as follows:
beancounter addportfolio SUNW:100:USD:bob:401k:85.50:19991117 \
IBM:100:USD:bob:401k:90.25:20000320
Here we specify that 100 shares each of Sun and IBM, priced in US Dollars, are in Bob's portfolio which is tagged as a 401k retirement account. The (fictitious) purchase price and date are also given.
Let's suppose that Bill prefers networking equipment, and that he has a brokerage account in Canada:
beancounter addportfolio CSCO:100:USD:bill:spec:78.00:19990817 \
NT:200:CAD:bill:spec:cad:90.25:20000212
Now we can backpopulate the database from 1998 onwards for all four stocks:
beancounter backpopulate --prevdate 19980101 CSCO IBM NT SUNW
With this historical data in place, we now compare how Bob's portfolio would have fared over the last 18 months:
beancounter plreport --prevdate '18 months ago' \
--restriction "owner='bob'"
Note how we use double quotes to protect the arguments, and how the SQL restriction contains a further single quote around the literal string.
We can also review the performance for Bill at the most recent trading day:
beancounter dayendreport --restriction "owner='bill'"
or the status of holdings and their respective values:
beancounter dayendreport --restriction "owner='bill'"
Similarly, a risk reports can be run on this portfolios per
beancounter risk --restriction "owner='bill'"
addportfolio is the most important 'position entry' command. As with other commands, several arguments can be given at the same time. For each of these, records are separated using a colon and specify, in order, stock symbol, number of stocks held, currency, account type, account owner, purchase price and purchase date. Only the first three arguments are required, the others are optional. Executing addportfolio implicitly executes addstock. The account type column can be used to specify whether the account is, e.g., a tax-sheltered retirement account, or it could be used to denote the brokerage company is it held at.
plreport retrieves the most recent quotes(s). This is useful for illiquid securities which might not have traded that day, or if a public holiday occurred, or if there was a data error at Yahoo!. Two dates can be specified which determine the period over which the profit or loss is computed. This will fail if price data (or currency data in the case of foreign stocks data) data is not available for either of those two dates. This can be restrictive for foreign stocks where we cannot backpopulate due to lack of public data source for historical currency quotes.
dayendreport is similar to plreport but is always over a one-day period. It also uses only one date record by calculating performance given the 'previous close' data.
status shows holdings amounts, total position values, annualized returns in percentages and holding periods in days. Note that the annualized returns can appear excessive if, e.g., a ten-day return from a recently purchased stock is extrapolated to an annual time period.
risk shows a portfolio risk report which describes the
statistically plausible loss which should be exceeded only 1 out of
100 times. In other words, the loss estimate has a critical level of
99%. This risk level is estimated via two methods. The first is
non-parametric and assumes no particular model or distribution; it
computes the 1% quintile of the return distribution and displays it as
well as the corresponding asset value at risk. The second method uses
the standard Value-at-Risk (VaR) approach. This uses the 1% critical
value of the Normal distribution and implicitly assumes a normal
distribution for returns. See http://www.gloriamundi.org for more
introduction and references. If the distribution of normalitty was
perfectly true, both measures would coincide. A large difference
between the two estimates would indicate that the return distribution
might be rather non-normal. Another view of the riskiness of a given
position is provided by the last column with the 'margVaR' heading. It
shows the marginal Value-at-Risk. Marginal VaR is commonly defined as
the risk contribution of the given position to the total portfolio,
and calculated as the difference in the VaR of the full portfolio and
the VaR of an otherwise identical portfolio with the given position
removed. Note that calculating marginal VaR is fairly slow (on the
order of O(n^3) ].
retracement shows a 'drawdown' report. Drawdown is commonly defined as the percentage loss relative to the previous high. The default period is used, but can be altered with the --date and --prevdate options. The default period is also corrected for the actual holding period. In other words, if a stock has been held for two months, only those two months are used instead of the default of six months -- but if the last months has been selected via --prevdate then it is used. For short positions, the analysis is inverted and relative to the previous low. The report displays each stock, the number of shares held, the current price and holdings value. The next two columns show the maximum price attained in the examined period, and the percent decline relative to it. The last column shows the unrealized loss relative to the maximum price over the period. The aggregate holdings value, percent decline and unrealized loss are shown as well.
advancement does the opposite of drawdown -- it computes unrealized gains relative to the minimum price in the period. The discussion in the preceding paragraph applies `but inverted'.
addindex adds stocks a the index table. Currently, no further analysis references this table.
addstock adds stocks to the database. From then on data will be
retrieved for the given symbol(s) and stored in the database whenever
the update command is executed.
addpopulate fills the database with historic prices for the given symbols and date period. Note that this works for stocks only, but not for currencies due to the lack of a public source of historical FX prices.
quote simply shows a price quote for the given symbol(s).
update updates the database with quotes for all stocks for the given day. No output is generated making the command suitable for cron execution.
dailyjob is a simple convenience wrapper around update, dayendreport, status and risk,
allreports is a another covenience wrapper around dayendreport, status and risk.
delete removes the given symbols from the database.
destroydb deletes the BeanCounter database.
warranty display a short GNU General Public License statement.
--currency can be used to select a different home currency. Instead of having all values converted to the default currency, the selected currency is used.
--date allows to choose a different reference date. This is then be be used by commands working on a date, or date period, such as plreport, dayendreport, backpopulate or status. --prevdate allows to choose a different start date for return calculations, or data gathering.
--restriction can be used to restrict the database selection. The argument must be a valid part of valid SQL statement in the sense that existing columns and operators have to be employed. The argument to this option will be completed with a leading and. The SQL restriction will typcally be over elements of the portfolio table which comprises the columns symbol, shares, currency, type, owner, cost and date. A simple example would be currency='CAD'. Note that this has to protected by double quotes ``I on the command-line.
--extrafx allows to gather data on additional currency rates beyond those automatically selected as shares are listed in them. A typical example would be for a European investor wanting to convert from the EUR in which the shares are listed into one of the member currencies which beancounter would no longer retrieve as shares are no longer listed in these.
--forceupdate allows to overwrite an potentially wrong date in the database update. Unfortunately, it appears that Yahoo! occasionally reports correct prices with an incorrect date such as the previous day's. In such a case, this option, along with an argument such as 'today' can override the bad date datapoint and avoid a hole in the database. The downside of this approach is that it would ``double'' the previous data in the case of a public holiday, or even if it was run the weekend. A somewhat smarter comparison to previously stored data might prevent that, but would be more complex to implement.
--rcfile allows to specify a resource file different from the default ~/.beancounterrc.
--dbsystem allows to switch to a different database backend. The default is PostgreSQL but MySQL is also supported.
--dbsystem allows to switch to an alternate database. The default is 'beancounter'. This can be useful for testing new features.
--fxupdate is a boolean switch to enforece updates of FX rates during 'update'. The default is 'true' but '--nofxupdate' can be used to suppress the update of foreign exchange rates.
The --verbose and --debug switches can be used in debugging an testing, and --help triggers the display of help message.
The following section details some of the database and configuration options.
beancounter currently depends on either PostgreSQL or MySQL, or any other database for which an ODBC driver is available (though the required tables would have to created manually in the ODBC case). Yet another DB backend could be added provided suitable Perl DBI drivers are available. For PostgreSQL and MySQL, the setup_beancounter script can create and initialize the database, form the required tables and fills them with some example data. It is a starting point for local modifications.
The connection to the database is made via a dedicated function in the BeanCounter.pm module, changes would only have to be made there. As of this writing the Perl DBI (the database-independent interface for Perl) is used along the DBI drivers for PostgreSQL, MySQL and ODBC. Ports for MySQL, Oracle, ... are encouraged.
A configuration file ~/.beancounterrc is read if found. It currently supports the following options:
There are now several ODBC systems available for Linux / Unix. The following ~/.odbc.ini work with the iODBC library and the PostgreSQL ODBC driver on my Debian GNU/Linux system:
[ODBC Data Sources] beancounter = BeanCounter Database
[beancounter] Driver = /usr/lib/libpsqlodbc.so Database = beancounter Servername = localhost
[ODBC] InstallDir = /usr/lib
Suggestions are welcome of how to set beancounter up with other ODBC libraries.
Finance::BeanCounter and beancounter are so fresh that there are only missing features :) Seriously, check the TODO list. This code or its predecessors have been used by the author since the end of 1998.
Finance::BeanCounter.3pm, smtm.1, Finance::YahooQuote.3pm, LWP.3pm, Date::Manip.3pm, Statistics::Descriptive.3pm
beancounter is (c) 2000, 2001, 2002 by Dirk Eddelbuettel <edd@debian.org>
Updates to this program might appear at http://eddelbuettel.com/dirk/code/beancounter.html.
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. There is NO warranty whatsoever.
The information that you obtain with this program may be copyrighted by Yahoo! Inc., and is governed by their usage license. See http://www.yahoo.com/docs/info/gen_disclaimer.html for more information.
The Finance::YahooQuote module by Dj Padzensky (on the web at http://www.padz.net/~djpadz/YahooQuote/) served as the backbone for data retrieval, and a guideline for the extension to the non-North American quotes which was already very useful for the real-time ticker http://eddelbuettel.com/dirk/code/smtm.html.