No trades for EuroStoxx50 ; consistent trades for Google (same code ; all data from yfinance). Any idea?

[nicolas-ghub]

I start with backtesting.py and I’m lost.

I have the following code to test a very simple strategy as a first program (Buy when MACD becomes positive; Sell when MACD becomes negative).

The code gives consistent results for Google but I don’t get any transactions for EuroStoxx50 while the conditions are met (see 'trades_control' DataFrame: there should be a purchase on 2010-06-01 for example).

The format of dates and data seems identical between Google and Eurostoxx so I have no explanation for this difference in result.

Does anyone understand what the problem is?

Here is the code:

#!/usr/bin/env python3

import pandas as pd
import pandas_ta as ta
import numpy as np

import yfinance as yf

import backtesting
from backtesting import Backtest
from backtesting import Strategy
from backtesting.lib import crossover
from backtesting.test import GOOG

Step 1. Data

# . E : Euro Stoxx 50
# . G : Google
Data_choice = 'E'

if Data_choice == "E":
ESTX50 = yf.download('^STOXX50E',
start="2010-01-01",
end="2024-02-07",
progress=True)

ESTX50.drop("Adj Close", axis=1, inplace=True)   
ESTX50.to_excel("ESTX50.xlsx", sheet_name='ESTX50_yfinance', index=True)   
data = ESTX50

elif Data_choice == "G":
# data = GOOG
Google = yf.download('GOOG',
start="2010-01-01",
end="2024-02-07",
progress=True)

Google.to_excel("Google.xlsx", sheet_name='Google_yfinance', index=True)    
data = Google

Step 2. Strategy definition

class MACD_crossing_zero(Strategy):

fast = 12
slow = 26
signal = 9
macd_level = 0

# 2a. Strategy definition: indicator
# ta.macd (pandas TA). Outputs pd.DataFrame: macd, histogram, signal columns.
def init(self):
    self.macdl, self.macdh, self.macds = self.I(ta.macd, pd.Series(
        self.data.Close), self.fast, self.slow, self.signal)

# 2b. Strategy definition: decision rule
#     Crossover: Return True if series1 just crossed over (above) series2.
def next(self):
    if crossover(self.macdh, self.macd_level):
        self.buy()
    elif crossover(self.macd_level, self.macdh):
        self.position.close()

Step 3. Backtest parameters

bt = Backtest(data, MACD_crossing_zero, cash=1_000, commission=.001,
trade_on_close=False, exclusive_orders=True)

Step 4. Backtest running

stats = bt.run()

trades = stats['_trades'].to_string()

Step 5. Backtest control

macd_control = ta.macd(data.Close, fast=12, slow=26, signal=9)

macdl = macd_control.iloc[:, 0]
macdh = macd_control.iloc[:, 1]
macds = macd_control.iloc[:, 2]

# Concatenate MACD histogram and Trades in the same pd DataFrame
# (ie with the same calendar to check decision consistency)
trades_control0 = pd.DataFrame(stats['_trades'])
trades_control0.set_index('EntryTime', inplace=True)

trades_control = pd.concat([macdh,trades_control0], axis=1)

trades_control = trades_control.fillna('')
trades_control = trades_control.replace({np.nan: None})
trades_control = trades_control.replace({np.nan: ""})

=> Visual inspection at the end:

. trades

. trades_control

[nicolas-ghub]

Problem solved.
Backtesting.py was explicit ... it was enough to read the warning ...

UserWarning: Some prices are larger than initial cash value. Note that fractional trading is not supported. If you want to trade Bitcoin, increase initial cash, or trade μBTC or satoshis instead (GH-134)."