On Options Site

People in India do not have good knowledge of options. Scope seems a lot. My father doesn't even understand delta and arbitrage. How can these people trade options without guidance? These people shouldn't be allowed to trade without education. Simple put call parity is not understood by most of the people.

Implemented a basic svm non linear classifier for predicting the direction. Saw an absolutely stupid implementation of linear regression for stock price prediction. How can people be so stupid? Such stupidity must be criminalized. Made me question if I understood it right. I guess good feature engineering should work. I do think that there are patterns in the prices and that stock prices are not really a random walk.

As usual, deep learning might not really be necessary. Good feature engineering with SVM should be enough.

Will have to brush up a little bit. Will have to go deeper to make it useful from a commercial point of view.

Will backtest put call parity extensively. May be a small fortune is hiding there somewhere. I am certain that temporary mispricing during fast moving markets, especially in far month options exist.

I think separate models will be needed for different timeframes. Also maybe different phenomenas will probably be caught better by different models. Will try single models as well but I think individual models will capture it better.

Rather than building one single model, lets see if an ensemble approach gives a better result.  I think some people in ML industry do not use the concept of PSI. This can translate into poor performance for the model in real world if the training data and real data differ significantly.  This issue can perhaps be avoided or reduced using better algorithms and techniques.

Stock market prediction is perhaps one of the hardest problems in data science I think a basic real world workable stock market model is going to take at least a month.  And I am talking about a very basic but workable model.

Charging hacked. Cant work.

Implemented a basic HMM. Will try different things to see what works best.
Starting cycles analysis. Area under the curve should give degree of expansion and cooling. Perhaps reduced need to consider amplitude and duration.

Patterns also emerge because information is processed gradually by the market. It isn't processed in one go as suggested by EMH. Human psychology and economic and business cycles also play a major role in pattern formation.