Home Archive Trilium Notes About

Net worth tracking

Posted on 2015-03-02

I like to keep a close track on my spending and income and to optimize both using historical data. I used to have a large spreadsheet for this purpose, but I was lucky to find the excellent HomeBank instead of keeping it. Unfortunately, HomeBank alone eventually wasn’t enough to track all my assets.

I have a custom setup which collects the values of my assets. Most of the code is at github.com/MichalPokorny/scripts (with horrible undocumented dependencies on btckit, worthy and maybe a few others). In this post, I give a high-level overview of how it works at the moment.

The problem

The system needs to keep track of various assets which aren’t straightforward to appraise:

Bitcoin price 06/2014-06/2015 - the largest Bitcoin bubble yet

Bitcoin price 06/2014-06/2015 - the largest Bitcoin bubble yet. The price since dropped down to today’s ~$250. Image from bitcoincharts.com.

The original problem was keeping track of my fiat money and of my Bitcoins. I wanted to have something stupidly hacky and working. Net worth tracking may have better solutions today (even for the weird structure of my assets), but I guess I’m going to keep using this solution until I’m bored enough to find a replacement :)

Collecting and storing the data

I have a Ruby script that figures the value of my assets using various sources and then saves the data. The script is in ~prvak/bin/net-worth-keeper and it’s run as a cronjob every hour or so:

# crontab
0 */1 * * * net-worth-keeper --log --quiet

Cron has the added benefit of sending me a mail when the script fails.

I keep all the data in a simple CSV file at ~prvak/misc/net-worth.csv. CSVs are simple and readable by everything (think OpenOffice Calc, Gnuplot, Ruby, Octave, etc.). They are also human-readable and writeable.

The beginning of the CSV file looks like this (data obviously scrambled):

2013-09-28 21:00:05,1380394805,11013579.15,123456.7,890123.45,9999999,0.0,0.0
2013-09-28 21:16:15,1380395775,11015200.366,125000.7,890200.666,9999999,0.0,0.0

Note the zeroes at the end - for example, I didn’t have any stocks at first. When I needed to start tracking them, I just wrote the tracking code and ran a complicated NoSQL CSV migration (i.e. I added “,0.0” at the end of each line).

New entries are always appended at the end of the file.

The sources of asset values

The original net-worth-keeper just fired a bunch of other programs using the parallel gem. Every program’s responsibility was to get the price of one asset (e.g. my Bitcoins) and output it in CZK. net-worth-keeper waited until those programs crashed or gave a valid output, parsed the output and appended it to the CSV file.

Those binaries shared a lot of common code or common libraries (like currency conversion) and they fired a lot of expensive Ruby interpreters. I eventually moved shared code into one directory. Now net-worth-keeper calls a library function (Prvak::Finance::NetWorth.load_assets), which returns a breakdown of asset values as a Ruby Hash (like {'Money' => 12345, ...}).

The breakdown is still done in parallel, but now most of the “workers” are actually just Ruby functions (like Prvak::Finance::IKSPortfolio.load.value or Prvak::Finance::Homebank::Accounting.load.total_value). I can still use separate executables where needed. Every data source just returns a single number (of CZK) or crashes and burns.

I have the following data sources:

Using the data

The net-worth-keeper binary by default runs in an “interactive mode”, which asks the the terminal-table gem to print a summary of asset values (remember, I track each asset separately, not just the sum of everything). In addition to this, I also have a small Gnuplot file in ~prvak/misc/plot-net-worth.gnuplot, which plots a stacked plot of asset values:

set datafile separator ','
set xdata time
set format x "%d.%m"
set timefmt "%Y-%m-%d %H:%M:%S"
set key vert out center left
set autoscale xfixmin

plot \
     "net-worth.csv" u 1:($6+$5+$4+$7+$8) w filledcurves x1 title 'EUR účet', \
     "" u 1:($6+$5+$4+$7) w filledcurves x1 title 'Akcie', \
     "" u 1:($6+$5+$4) w filledcurves x1 title 'KB účty', \
     "" u 1:($6+$5) w filledcurves x1 title 'IKS fondy', \
     "" u 1:6 w filledcurves x1 title 'Bitcoiny', \
     "" u 1:6 w lines title 'Bitcoiny', \
     "" u 1:5 w lines title 'IKS fondy', \
     "" u 1:4 w lines title 'KB účty', \
     "" u 1:7 w lines title 'Akcie', \
     "" u 1:8 w lines title 'EUR účet', \

I run it using in an interactive Gnuplot session. Sorry for not including the output - I’m not comfortable with giving it to everyone on the internet. If you squint just right, you can actually see some trends where they should be :)

Final notes

I have recently started rewriting parts of the code to Go. There are many things I like about Go - like the syntax, the blazing speed of tools, the One True Formatting (gofmt), or the package management (imports + go get). On the other hand, it is much less annoying to write quick JSON/YAML parsers in Ruby than in Go, but I hope this is just a temporary rough edge. The language is also less mature, so some libraries don’t quite exist yet (like Mechanize). However, I have high hopes for Go.

It would be nice to extend this system one day to track my “savings ratio”. The “savings ratio” is exactly what it says on the tin - how much money you save divided by how much money you earn (not including investment income). I hope to write one day about why I think tracking this value would be useful.

This system has problems when I’m moving larger amounts of money through “asset boundaries” - for example, it takes some time for money to travel between bank accounts. One needs to remember to put the transaction in both places once it finishes. A similar problem arises when I expect to be reimbursed for an expense by an employer - I don’t want to record that as a wedge in my nice graphs, but I also don’t want to have large disagreements between my records and the real state of my accounts. I don’t think I can have both.