Binance Bot Trader and Simulator for Random Forest and other strategies RSI, MACD, Bollinger Bands…

ALAgrApHY
4 min readApr 2, 2024

First and foremost, I am not a financial advisor or a professional trader and you are responsible from any investments made, however, I have long been interested in understanding the market movement and whether an investment in it is worthwhile.

I tried several algorithms (Random Forest, LSTM) and strategies (RSI, MACD, Bollinger Bands and combinations thereof with various parameters) that you can simply run as a trading bot or as a simulator to have personal insights on specific stocks since my choice of stocks/cyptos might not have been optimal.

My open source code works for any account on the largest crypto trading platform, binance and all you need to get started is your public and private API with some of the most commonly used parameters for RF, MACD, BB and RSI.

from binance.client import Client
import ta
import time
import pandas as pd
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error
import numpy as np

# Initialize Binance client
client = Client(api_key, api_secret)

# Trading pair and parameters
import sys
trading_pair = 'SOLUSDT' #change this to whatever crypto or stock you are interested in trading
btc_pair = 'BTCUSDT'

ass = trading_pair[:-4]

bb_window = 21
bb_std_dev = 2
macd_slow = 26
macd_fast = 12
macd_signal = 9
rsi_period = 14
rsi_overbought = 70
rsi_oversold = 30
rf_window = 60 # Window size for Random Forest

A couple of simple functions allow your algorithm to trade on your behalf but these are commented by default to make sure you run the simulator a few times before deciding to actually trade:


# Function to get the current price
def get_current_price(symbol):
ticker = client.get_ticker(symbol=symbol)
return float(ticker['lastPrice'])

# Function to place a buy order
def buy(symbol, quantity):
order = client.order_market_buy(symbol=symbol, quantity=quantity)
print(f"Buy order placed for {symbol}: {order}")

# Function to place a sell order
def sell(symbol, quantity):
order = client.order_market_sell(symbol=symbol, quantity=quantity)
print(f"Sell order placed for {symbol}: {order}")

Next, we introduce an infinite loop that will calculate RFI, MACD and BB values from klines of the past 1 minute (that can be set to a larger frame for less frequent trades) while the Random Forest will be trained on the past klines data to predict a forecast.

# Main trading loop
while True:
# Get historical OHLCV data
klines = client.get_historical_klines(symbol=trading_pair, interval=Client.KLINE_INTERVAL_1MINUTE)
btc_klines = client.get_historical_klines(symbol=btc_pair, interval=Client.KLINE_INTERVAL_1MINUTE)


# Get current price
current_price = get_current_price(trading_pair)

# Calculate RSI
closes = [float(kline[4]) for kline in klines]
btc_closes = [float(kline[4]) for kline in btc_klines]



closes_series = pd.Series(closes) # Convert list to pandas Series
btc_closes_series = pd.Series(btc_closes)

rsi_values = ta.momentum.rsi(closes_series, window=rsi_period)
current_rsi = rsi_values.iloc[-1] # Get the last RSI value
print("## RSI current price and RSIi/overbought", current_price, rsi_oversold,"<",current_rsi,"<",rsi_overbought)

# Calculate Bollinger Bands
bb_indicator = ta.volatility.BollingerBands(closes_series, window=bb_window, window_dev=bb_std_dev)
bb_lower = bb_indicator.bollinger_lband().iloc[-1]
bb_upper = bb_indicator.bollinger_hband().iloc[-1]

# Calculate MACD
macd = ta.trend.MACD(closes_series)
macd_line = macd.macd().iloc[-1]
macd_signal_line = macd.macd_signal().iloc[-1]


# Prepare data for Random Forest
X = []
y = []
for i in range(rf_window, len(closes)):
X.append(list(closes[i - rf_window:i])+ list(btc_closes[i - rf_window:i]))
y.append(closes[i])

X = np.array(X) # Convert list to NumPy array

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train Random Forest Regressor
rf_model = RandomForestRegressor(n_estimators=100, random_state=42)
rf_model.fit(X_train, y_train)

# Predict the next price using Random Forest
next_price_pred = rf_model.predict([X_test[-1]])
print("## RF current price and predicted price", current_price, next_price_pred)




# Check trading conditions
if current_rsi > rsi_overbought:
# Sell condition
balance = client.get_asset_balance(asset=ass)
sell_quantity = float(balance['free'])
if sell_quantity > 0:
print("\nRSI sell", sell_quantity )
#sell(trading_pair, sell_quantity)
elif current_rsi < rsi_oversold:
# Buy condition
usdt_balance = client.get_asset_balance(asset='USDT')
buy_amount = float(usdt_balance['free'])
if buy_amount > 0:
buy_quantity = buy_amount / current_price
print("\nRSI buy amount",buy_amount, "price", current_price, "total:",(buy_amount / current_price))
#buy(trading_pair, buy_quantity)
print("## BOL",bb_upper,"<",current_price,"<", bb_lower)
print("## MACD MACD_line:", macd_line, ">" ,macd_signal_line)
# Check trading conditions
if current_price < bb_lower and macd_line > macd_signal_line:
# Buy condition
usdt_balance = client.get_asset_balance(asset='USDT')
buy_amount = float(usdt_balance['free'])
if buy_amount > 0:
buy_quantity = buy_amount / current_price
print("\nMACD/BOL buy amount",buy_amount, "price", current_price, "total:",(buy_amount / current_price))
#buy(trading_pair, buy_quantity)
elif current_price > bb_upper and macd_line < macd_signal_line:
# Sell condition
balance = client.get_asset_balance(asset=ass)
sell_quantity = float(balance['free'])
if sell_quantity > 0:
#sell(trading_pair, sell_quantity)
print("\nMACD/BOL sell", sell_quantity )

print("---------------------")
# Wait for the next iteration
time.sleep(60) # Sleep for 1 minute

You can either uncomment the buy/sell calls or based on the printouts do it manually. You can also play with the default parameters and add in a few more functions.

One of the functions I have also tested is the ARIMA after installing pmdarima and importing autoarima.

from pmdarima.arima import auto_arima

# ARIMA
arima_model = auto_arima(train_data['Price'], trace=True, error_action='ignore', suppress_warnings=True)
arima_predictions = arima_model.predict(n_periods=len(test_data))
arima_mse = mean_squared_error(test_data['Price'], arima_predictions)
arima_accuracy = 1 - (arima_mse / test_data['Price'].var())

In another post, I am excited to reveal the result of having simulated some of the top stocks with all trading fees, tax application for a net profit calculation to finally answer one big question:

Is trading (long/short term) worth it?

Bull and bear market laughing at every trader — crypto trading algorithm — trading bot

I guess you’ll have to follow my posts to find out. I will also announce it on my bitopsy youtube channel.

For the entire binance crypto bot trading code: https://github.com/bitopsy/Binance-crypto-bot-trader-multi-strategy/tree/main

--

--

ALAgrApHY

Traditional & AI Artist, Data Scientist, Filmmaker exploring Creative AI. Started the GAN AI Art Movement in 2016. Former Postdoc @CNRS & PhD @INFORMATICS