Live Data/Trading Not Respecting Timeframe/Compression

monster_wig

I am using the Alpaca paper account as my broker and testing a very basic scenario just to validate functionality. I have been unable to implement trades based on the 1 minute, 5 minute or 1 hour bar. It seems that the strategy just attempts to trade off the real-time second-by-second data instead of the 1/5/60 minute bars.

Since I am using Alpaca I use the "alpaca_backtrader_api" python library along with the backtrader library. The following code produces the second-by-second "next" output.

import config
import alpaca_backtrader_api
import backtrader as bt
from datetime import datetime

ALPACA_API_KEY = config.alpaca_paper_api_key_id
ALPACA_SECRET_KEY = config.alpaca_paper_api_secret_key
ALPACA_PAPER = True


class EmptyStrategy(bt.Strategy):

    def __init__(self):
        print(f"Define Empty Strategy")

    def next(self):
        date_fmt = '%Y%m%d %H:%M:%S'
        date_str = self.data.datetime.datetime(0).strftime(date_fmt)
        print(f"{date_str}: O:{self.data.open[0]:.2f} C:{self.data.close[0]:.2f}")


cerebro = bt.Cerebro()

store = alpaca_backtrader_api.AlpacaStore(
    key_id=ALPACA_API_KEY,
    secret_key=ALPACA_SECRET_KEY,
    paper=ALPACA_PAPER,
    usePolygon=False,
)

alpaca_data_kwargs = dict(
    dataname='AAPL',
    timeframe=bt.TimeFrame.Minutes,
    compression=5,
    historical=False,
    fromdate=datetime.now()
)

broker = alpaca_backtrader_api.AlpacaBroker()

# DataFactory = alpaca_backtrader_api.AlpacaData
DataFactory = store.getdata
data0 = DataFactory(**alpaca_data_kwargs)

cerebro.addstrategy(EmptyStrategy)
cerebro.setbroker(broker)
cerebro.addstore(store)
cerebro.adddata(data0)

print(f"Running Cerebro...")
cerebro.run()

Output is:

Running Cerebro...
Define Empty Strategy
20201006 15:01:26: O:9.25 C:9.25
20201006 15:01:28: O:9.22 C:9.22
20201006 15:01:31: O:9.22 C:9.22
20201006 15:01:34: O:9.22 C:9.22
20201006 15:01:41: O:9.22 C:9.22
20201006 15:01:44: O:9.22 C:9.22
20201006 15:01:45: O:9.25 C:9.25
20201006 15:01:47: O:9.25 C:9.25
20201006 15:01:51: O:9.22 C:9.22
20201006 15:01:52: O:9.27 C:9.27
20201006 15:01:53: O:9.27 C:9.27

The output shows the "next" method is getting called every second or so which does not make sense to me when I specified "timeframe=bt.TimeFrame.Minutes" and "compression=5". I would have expected to only see output every 5 minutes.

When I ran an actual strategy it was clear to me trades were occurring based on the second-by-second data, which is not what I wanted.

Am I doing something wrong in my example?

ab_trader

i would ask this question to alpaca data feed developers.