Zscore_Crypto / zscore_backtest.py
gjin10969
initialize
e97cf97
raw
history blame
8.43 kB
import pandas as pd
import ccxt
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import json
from datetime import datetime, timedelta
import pytz
import os
images_folder = '/home/gjin/Documents/zscore/images'
os.makedirs(images_folder, exist_ok=True)
# Prompt for the symbol and time frame
symbols = input('Please input Symbol: ')
timeframe = input("Please input time frame: ")
# Initialize Binance Futures API
binance = ccxt.binance({
'options': {'defaultType': 'future'}, # Specify futures
})
from tqdm import tqdm
import pandas as pd
def fetch_and_calculate_zscore(symbol, timeframe, since, limit=200, rolling_window=30):
data = binance.fetch_ohlcv(symbol, timeframe=timeframe, since=since, limit=limit)
df = pd.DataFrame(data, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
# Convert timestamp to UTC datetime format
df['timestamp'] = pd.to_datetime(df['timestamp'], unit='ms', utc=True)
# Calculate rolling mean, std, and Z-Score
df['mean'] = df['close'].rolling(window=rolling_window).mean()
df['std'] = df['close'].rolling(window=rolling_window).std()
df['z_score'] = (df['close'] - df['mean']) / df['std']
# Initialize signal columns
df['buy_signal'] = 0
df['sell_signal'] = 0
# Variables to track thresholds and state
in_sell_signal = False
in_buy_signal = False
crossed_threshold = False # Track if the extreme threshold has been crossed (2 or -2)
# Iterate through the dataframe to track signals with tqdm progress bar
for i in tqdm(range(1, len(df)), desc="Processing Z-score signals"):
current_z = df.loc[i, 'z_score']
# Track when Z-score crosses the thresholds (2 and -2)
if current_z > 2 and not crossed_threshold: # If Z-score exceeds 2
crossed_threshold = True
in_sell_signal = True # Trigger sell signal
elif current_z < -2 and not crossed_threshold: # If Z-score falls below -2
crossed_threshold = True
in_buy_signal = True # Trigger buy signal
# Maintain sell signal between 1 and 2
if in_sell_signal:
if 1 <= current_z <= 2:
df.loc[i, 'sell_signal'] = 1 # Sell signal active
# Exit sell signal if Z-score falls below 1
elif current_z < 1:
in_sell_signal = False
crossed_threshold = False # Reset threshold crossing
# Maintain buy signal between -2 and -1
if in_buy_signal:
if -2 <= current_z <= -1:
df.loc[i, 'buy_signal'] = 1 # Buy signal active
# Exit buy signal if Z-score rises above -1
elif current_z > -1:
in_buy_signal = False
crossed_threshold = False # Reset threshold crossing
return df
# Convert time to local timezone (Philippine Time)
utc_time = datetime.utcnow()
philippine_tz = pytz.timezone('Asia/Manila')
philippine_time = pytz.utc.localize(utc_time).astimezone(philippine_tz)
# Format the time in your preferred format
formatted_ph_time = philippine_time.strftime("%Y-%m-%d %H:%M:%S")
latest_data = []
def update_signal_json(symbol, df, json_data):
# Extract the latest data point from the DataFrame
global latest_data
latest_data = df.iloc[-1] # Update to get the last row
# Get the timestamp from the latest data and format it
timestamp = latest_data['timestamp'].strftime("%Y-%m-%d %H:%M:%S") if isinstance(latest_data['timestamp'], pd.Timestamp) else str(latest_data['timestamp'])
# Check if the latest Z-score has a signal
signal_status = "True" if latest_data['buy_signal'] == 1 or latest_data['sell_signal'] == 1 else "False"
# Prepare new entry with real-time Z-Score
signal_entry = {
"symbol": symbol,
"time_frame": timeframe, # Make sure `timeframe` is defined or passed to this function
"date_and_time": timestamp, # Correct timestamp for the entry
"realtime_ph_time": formatted_ph_time, # Add the local Philippine time (UTC+8)
"current_price": latest_data['close'], # Closing price for the most recent entry
"zscore": latest_data['z_score'], # Z-Score value
"detection": signal_status # Add signal status
}
# Append the new data to the json_data list
json_data.append(signal_entry)
return json_data
def plot_data(btcdom_df, pair_df, btc_df):
fig, ax = plt.subplots(figsize=(14, 7))
# Clear previous plots
ax.clear()
# Plot Z-Scores for all pairs
ax.plot(btcdom_df['timestamp'], btcdom_df['z_score'], label="BTCDOM/USDT Z-Score", color='blue', linestyle='-')
ax.plot(pair_df['timestamp'], pair_df['z_score'], label=f"{symbols}/USDT Z-Score", color='orange', linestyle='-')
ax.plot(btc_df['timestamp'], btc_df['z_score'], label="BTC/USDT Z-Score", color='gray', linestyle='-')
# Add thresholds
ax.axhline(y=2, color='red', linestyle='--', label='Overbought Threshold')
ax.axhline(y=-2, color='green', linestyle='--', label='Oversold Threshold')
# Plot Buy and Sell signals for BTCDOM/USDT
ax.scatter(btcdom_df[btcdom_df['buy_signal'] == 1]['timestamp'], btcdom_df[btcdom_df['buy_signal'] == 1]['z_score'],
marker='^', color='green', label='BTCDOM Buy Signal')
ax.scatter(btcdom_df[btcdom_df['sell_signal'] == 1]['timestamp'], btcdom_df[btcdom_df['sell_signal'] == 1]['z_score'],
marker='v', color='red', label='BTCDOM Sell Signal')
# Plot signals for the other pair
ax.scatter(pair_df[pair_df['buy_signal'] == 1]['timestamp'], pair_df[pair_df['buy_signal'] == 1]['z_score'],
marker='^', color='green', alpha=0.5, label=f"{symbols} Buy Signal")
ax.scatter(pair_df[pair_df['sell_signal'] == 1]['timestamp'], pair_df[pair_df['sell_signal'] == 1]['z_score'],
marker='v', color='red', alpha=0.5, label=f"{symbols} Sell Signal")
# Format plot
ax.set_title(f"Z-Scores Signals {timeframe} for {symbols}/USDT Futures", fontsize=16)
ax.set_xlabel("Time (UTC)", fontsize=12)
ax.set_ylabel("Z-Score", fontsize=12)
ax.xaxis.set_major_formatter(mdates.DateFormatter("%Y-%m-%d %H:%M"))
ax.legend(loc="upper left")
ax.grid(True)
plt.xticks(rotation=45)
# Save each plot with unique filename
global latest_data
plot_filename = f"/home/gjin/Documents/zscore/images/zscore_plot_{symbols}_{latest_data['timestamp'].strftime('%Y%m%d_%H%M%S')}.png"
plt.savefig(plot_filename)
plt.close(fig) # Close the figure to prevent memory issues
# plt.close(fig) # Close the figure to prevent memory issues
# Function to run historical data processing
def run_historical():
json_data = []
try:
with open(f'signals_{symbols}.json', 'r') as file:
json_data = json.load(file)
except FileNotFoundError:
pass
# Set start and end dates for the loop
start_date = datetime(2024, 9, 1)
end_date = datetime(2024, 11, 27)
# Loop through each month in the date range (or week, depending on your choice)
current_date = start_date
while current_date < end_date:
# Set 'since' to the start of each month or week (whichever you prefer)
since = binance.parse8601(current_date.strftime('%Y-%m-%dT%H:%M:%SZ'))
btcdom_symbol = 'BTCDOM/USDT'
pair_symbol = f'{symbols}/USDT'
btc_symbol = 'BTC/USDT'
# Fetch and process data
btcdom_df = fetch_and_calculate_zscore(btcdom_symbol, timeframe, since)
pair_df = fetch_and_calculate_zscore(pair_symbol, timeframe, since)
btc_df = fetch_and_calculate_zscore(btc_symbol, timeframe, since)
# Update signals and append to JSON
json_data = update_signal_json(pair_symbol, pair_df, json_data)
json_data = update_signal_json(btc_symbol, btc_df, json_data)
json_data = update_signal_json(btcdom_symbol, btcdom_df, json_data)
# Save updated signals to JSON
with open(f'signals{symbols}.json', 'w') as file:
json.dump(json_data, file, indent=4)
# Plot the data and save each plot separately
plot_data(btcdom_df, pair_df, btc_df)
# Move to the next chunk (next month/week)
current_date += timedelta(hours=4)
# Start historical data processing
run_historical()