Hi Santosh Turamari  , 

Let's break down how a model is trained for stock predictions. During the training period, the model doesn’t just memorize the data; it learns patterns, trends, and correlations. Although the input data for predictions might be new or previously unseen, the model relies on the patterns it learned during training to make accurate predictions.

Your current setup is in "predictive mode" rather than "learning mode." This means the model uses pre-trained parameters to make predictions without updating them. The predictions are based on the statistical relationships the model captured during training. It does not need to learn about new stocks in real-time unless it's specifically designed to do so, such as in online learning models.

Even if the model hasn't seen those specific 100 data points before, it has encountered similar patterns during training, allowing it to make predictions.

Let's see a Analogy: Think of it as a person who has learned to predict the weather by studying historical data. Even if they encounter today's weather for the first time, they use their prior knowledge to estimate tomorrow’s conditions. They aren’t learning from today's data anew but rather applying what they've already understood.

The key takeaway is that the model can make accurate predictions because it generalizes from historical patterns captured during training. The 100 values act as inputs, and the model leverages its "knowledge" to predict the outcome.

I hope you now understand how it predicts new stocks without needing to train the model again. If you have any doubts, feel free to ask.