smartchg

🐍 DCA-based asset exchange algorithm.

Inspired by PAC (Pre-Authorized Contribution plan) and Smart PAC.

Installation

This utility is available as a Python package on PyPI:

python3 -mpip install smartchg

Usage

There are some files in the example directory of this repo that can be useful to demonstrate how this tool works, so let's change directory first:

cd example/

We need a Python virtual environment ("venv") with some packages to do the demonstration:

python3 -mvenv venv
venv/bin/python3 -mpip install -r requirements.txt

Note: we refer to the source asset with the generic ticker symbol SRC, and to the destination asset with DST.

Now we need to fetch data related to some asset. To do that, we can use https://github.com/dmotte/misc/blob/main/python-scripts/ohlcv-fetchers/yahoo-finance.py.

Note: in the following commands, replace the local path of the invoke.sh script with the correct one.

~/git/misc/python-scripts/ohlcv-fetchers/invoke.sh yahoo-finance '^GSPC' -i1d -d2020-01-01T00Z -f'{:.6f}' > ohlcv-SPX500.csv

Now that we have the data, we can compute the output data and values:

rate=$(tail -1 ohlcv-SPX500.csv | cut -d, -f6)
python3 -msmartchg -a.15 -r"$rate" -t1000 --fmt-src='{:.2f}' --fmt-dst='{:.4f}' --fmt-{rate,simil}='{:.6f}' {ohlcv,smartchg,values}-SPX500.csv
grep '^sugg_' values-SPX500.csv

Note: each output value and entry field is described with a comment in the compute_stuff function's code. You can search for the strings # - entry and # - values in the smartchg/cli.py file to get an overview.

And finally display some nice plots using the plots.py script (which uses the Plotly Python library):

venv/bin/python3 plots.py -ros {smartchg,values}-SPX500.csv

For more details on how to use this command, you can also refer to its help message (--help).

Algorithm

The algorithm is based on the assumption that one need to exchange some amount of one asset (e.g. EUR) for another asset (e.g. SPX500) regularly (e.g. once a month, every month). Also, they cannot decide when to make the exchange (e.g. always on the same day of the month). Therefore, the only controllable variable is "how much" to exchange, that is, the quantity of assets to be exchanged.

This algorithm tries to optimize the quantity of assets to be exchanged based on the trend of the historical exchange rate values, in order to determine whether the current exchange rate is convenient (and therefore we should exchange more) or unconvenient (and therefore we should exchange less). In one simple sentence, the motto is: "Buy more when low!"

Note: the following explanation is for illustration purposes only. Please refer to the Python code for any details.

Symbol	Variable
$a$	apy
$m$	multiplier
$t$	target
$n$	len(data)
$d_i$	entry['days']
$r_i$	entry['rate']
$p_i$	entry['pred']
$\delta_i$	entry['offset']
$s_i$	entry['simil']
$S$	sugg_src

First of all, we need to calculate the exchange rate prediction and offset:

$$ p_i = r_1 (1 + a)^{d_i/365} \quad \quad \delta_i = r_i - p_i $$

Then we need the mean and standard deviation of the offset values:

$$ \mu_\delta = \frac{1}{n} \sum_{i=1}^{n} \delta_i \quad \quad \sigma_\delta = \sqrt{\frac{1}{n-1} \sum_{i=1}^{n} \left( \delta_i - \mu_\delta \right) ^ 2} $$

The similarity values are calculated using the following formula:

$$ s_i = \frac{\delta_i - \mu_\delta}{2 \sigma_\delta} $$

Finally, we need to compute the suggested SRC amount, taking the multiplier value into account:

$$ S = t (1 - s_n m) $$

For example, if $t=500$, $s_n=-1$, and $m=0.10$, we have:

$$ S = 500 \cdot (1 - (-1) \cdot 0.10) = 500 \cdot 1.1 = 550 $$

Development

If you want to contribute to this project, you can install the package in editable mode:

python3 -mpip install -e . --user

This will just link the package to the original location, basically meaning any changes to the original package would reflect directly in your environment (source).

If you want to run the tests, you'll have to install the pytest package and then run:

python3 -mpytest test