Why would we want to predict when we have all data available to calculate 'T'?

[Vintelligence]

Just to exchange thoughts. We calculate the 'T' score based on data available in the dataframe. For the last time step, all this data to calculate 'T' is, I believe, available. So why would we want to predict 'T'? What is the added value?

[PyaeHeinTun]

I think even if you can compute T from historical data, you need to predict what T will be at the current moment to make decisions. So, by comparing with model predicted T with threshold value we can know what the model predict if it buy or sell. So, if you want to use this kind of method, finding the best optimal threshold is important.

[goranv90]

Did you try changing thresholds? Could anyone share some parameters if you found some good parameter set or anything to be changed in config to improve it?