Make DRY

[amustapha]

I think it might be a good idea to use dependency injection or a centralized config handler to minimize code repition, e.g We have to define data sources with a timeframe, then when using the datasource in a Strategy, the strategy would also have a timeframe which is defined then our backtest also uses a timeframe.

I think it might be a nice idea for a strategy to be time frame agnostic... this might keep the Strategy class open enough to have an optimizer... e.g, I can create a strategy and try to find on what time frame it would perform best. I can have a strategy and try to tweak the input parameters...

just ideas in my head

[MDUYN]

Hey @amustapha thanks!, could you provide some code outline how this would look. If I understand you correctly this would look along the line of:

from investing_algorithm_framework import create_app, PortfolioConfiguration, \
    TimeUnit, CCXTOHLCVMarketDataSource, Algorithm, \
    CCXTTickerMarketDataSource, MarketCredential

# New data on every 2h interval
bitvavo_btc_eur_ohlcv_2h = CCXTOHLCVMarketDataSource(
    identifier="BTC-ohlcv",
    market="BITVAVO",
    symbol="BTC/EUR",
    time_frame="2h",
    window_size=200
)

# New data on every 1h interval
bitvavo_btc_eur_ohlcv_1h = CCXTOHLCVMarketDataSource(
    identifier="BTC-ohlcv",
    market="BITVAVO",
    symbol="BTC/EUR",
    time_frame="1h",
    window_size=200
)

# Ticker data for orders, trades and positions
bitvavo_btc_eur_ticker = CCXTTickerMarketDataSource(
    identifier="BTC-ticker",
    market="BITVAVO",
    symbol="BTC/EUR",
)

... App creation here

# Without explicitly defining the interval when the strategy runs, it would run when there is new data
# available for bitvavo_btc_eur_ohlcv_2h and bitvavo_btc_eur_ohlcv_1h. Meaning that this function runs every 1h and 2h (basically every hour)
@algorithm.strategy()
def on_data(algorithm: Algorithm, market_data: dict):
    # Access the data sources with the indentifier
    polars_df = market_data["BTC-ohlcv"]

if __name__ == "__main__":
    app.run()

I see a couple of limitations here that we then should address:

• How do we schedule a strategy when the data is not based on OHLCV data. Say we create a data provider that runs on twitter data for example.
• How do we schedule an algorithm with multiple strategies that each run on different time frames? Say we have a strategy that runs every 2 hours and the other runs every 15 mins.