Love to see this! I'm glad our conversation in the last video had at least a bit of influence on this one :). Also loved that you used a rolling window forecast. Another step that could be interesting to take is using a variety of GARCH (T-GARCH, EGARCH, etc.) models to predict volatility of the growth rate and even better. Use it to measure VaR (value at risk). This incorporates risk management into your model which I think is crucial.

I'm still only learning data science in my spare time, but I love the layering of techniques :) It's also helping me develop my math skills when I try to replicate the models in C# and validate my work.. Your method of explaining these techniques is terrific, thanks so much :)

I love this series! Will you do a walkthrough the code like for episode 1?

Just amazing

These financial data projects are severely interesting

Awesome content man, keep it coming!

Great video! Thanks for posting! Have you considered to use the Sharpe Ratio criteria for choosing what % you allocate to each stock / instrument once you have selected them using the ARMA model?

awesome video, thank you!

Awesome video, I wondered if you would try a time series to make predictions. Looking forward to seeing the outcome!

Trying to beat the stock market is generally going to be hard, given the number of experts investing in it. It'd probably be a better idea to try the experiment for more niche markets (prediction markets, video games economies that allow you to trade in-game currency for actual money, ...) or markets where the number of experts is significantly lower (e.g. sports betting, which you can treat as a market by instead betting on the other team, when you would otherwise sell).

Additionally, would be interesting to pull in an unrelated data source, for example top 10 daily news stories and do some sentiment analysis. You could see if the normal news correlates to stock movements

*I Love Your Videos!*

I meant to ask, is there going to be a video, or a notebook or source code that goes into the implementation of this method?

Wouldn't P(CI INC) necessarily be _less_ than P(FIT)? Because you can't predict the CI interval if you can't even fit the data using the model, right? Maybe you meant for P(CI INC) to be a *conditional* probability, i.e. P(CI INC | FIT)?

What are you using to trade the stocks themselves?

How many iitans are watching this video ❤❤

Not the best idea from a financial point of view.