The Palantir Project is an attempt to aggregate and take advantage of the patterns of tradable assets in the context of auto-correlation. Palantir exploits the efficiency of the market in the market to try make a long or short prediction As if now Palantir has been optimized soley for bitcoin (for its high volatility) but will soon expand into all aspects of trade.
At it's core palantir combines techinal indicators from different time periods and settings to form predictions about markets in the near-future. Then the optimal decisions are made to maximize profit.
The preceding chart shows a simulation of a model trained on data from 2020-06-01 to 2020-06-20 and actively playing on unseen data from 2020-06-21 to 2020-06-30 with 1 bitcoin and 200USD. Based on it's predictions palantir shorts or holds to maintain constant profit-growth.
pip install palantir-cli
or for the latest dev version...
pip install git+git:https://github.com/nathanielCherian/palantir.git
Palantir's CLI interface can be activated from terminal with
$ palantir
Palantir will look for special identification files in the current directory before it loads the software. If these files do not exist or are not correct palantir will create them.
Arguments with a * are positional required args while ones with ** are optional
and must be entered in the form foo=bar. (similar to *args and **kwargs in python)
PALANTIR> help
PALANTIR> get-btc *[directory:str] **[days:50]
PALANTIR> init *[data-directory:str]
PALANTIR> backtest *[data-directory:str] **[save:str] **[fee:0.001] **[cash:400] **[bitcoin:1]
PALANTIR> rename *[new_name:str]
PALANTIR> exit
PALANTIR> clear
Palantir is still a WORK IN PROGRESS and will need more work before it has become a useful tool
-
Bitcoin trading fees: After applying the maximum fee (.30% per transaction) to trades the basic machine suffers greatly. To counteract this we have to find the equilibrium between minimizing trades and maximizing profits. Note: This will not be a problem in the palantir-stocks -
Need to work on the CLI
Please report any issues here. Thanks!
All contributions are very welcome.