I have a project in which im trying to predict future Crypto/Stock prices based on historic prices. I have made an ARIMA model for this exact purpose. though is seems like the prediction chart goes flat after about 3 or 4 days of future price prediction. Im a beginner in AI models in Python. I had speculated that the model may be overfitted, but after asking someone i know that is somewhat an expert in building AI models, he said that it is very rare for an ARIMA model to get overfitted (since its a Time Series model). he also said that if the model was overfitted the chart would look like a parabolic chart.

any help is appreciated

import yfinance as yf
df = yf.download('BTC-USD', period = '24MO')

import yfinance as yf
df = yf.download('ETH-USD')
print(df)

import numpy as np, pandas as pd, matplotlib.pyplot as plt
import math
from statsmodels.tsa.arima.model import ARIMA
from sklearn.metrics import mean_squared_error, mean_absolute_error
to_row = int(len(df)*0.9)
training_data = list(df[0:to_row]['Adj Close'])
testing_data = list(df[to_row:]['Adj Close'])+list(df[-61:]['Adj Close'])

model_predictions = []
model = ARIMA(training_data, order = (4,1,0))
model_fit = model.fit()

output = list(model_fit.forecast(steps=36))

PS , I also tried changing the training data timeframe which also was a dead end.
My code might have stupid mistakes as i said Im a beginner.

I also want to add that i have reviewed similar posts on stackoverflow and i do think the problem i encountered is somewhat different and i could not solve my problems by reading the answers on those posts.