Discussion of article "CatBoost machine learning algorithm from Yandex with no Python or R knowledge required" - page 4

 
yeah I figured it out and I was gona delete the last part of my comment.  Thanks. 
 
dustovshio #:
yeah I figured it out and I was gona delete the last part of my comment.  Thanks. 

Good luck!

 

If you were to apply this to a different EA then do you just apply the catboost model before the order is place and leave everything else the same, or do you need to modify the model_CB() ,  or copy_arhiv()?  It doesn't seem to be opening orders when the CB model is applied. 



 
dustovshio #:

If you were to apply this to a different EA then do you just apply the catboost model before the order is place and leave everything else the same, or do you need to modify the model_CB() ,  or copy_arhiv()?  It doesn't seem to be opening orders when the CB model is applied. 



You can add or change the input signal in the Signal() function.

Have you trained the CatBoost model?

If you did everything right, it should work.

 

Yeah I did train the model. so what if the EA closes, reduces, or reverses positions on the opposite signal do you want to filter those using the model. or simply filter out new orders from opening?  



 
dustovshio #:

Yeah I did train the model. so what if the EA closes, reduces, or reverses positions on the opposite signal do you want to filter those using the model. or simply filter out new orders from opening?  



I didn't understand the thought: " do you want to filter those using the model ".

With the help of the model, signals for opening a position are filtered in the article.

 
Aleksey Vyazmikin #:

I didn't understand the thought: " do you want to filter those using the model ".

With the help of the model, signals for opening a position are filtered in the article.

Your if your EA has an opposite signal it can close orders.  If at boost can theoretically reduce false signals.  They if the opposite signal closes orders, then catboost would reduce the false orders from closing and the result would be you leave orders open longer and greater profit is achieved.  For example. You place an order when your MA crosses.  your stoploss is 50 pips and TP is 50. However, there MA crosses back over before you it your SL or TP, and your EA is programed to close the order it his happens: this is referred to as close (or reduce, or reverse) on the opposite signal.  Now if that signal was a false alarm, then you closes your profit too early, when it could have gone up to your TP instead.  So catboost could have filtered a certain percentage of those false signals? This is my question.  Not all EA's close positions on the opposite signal. Many just have a fixed Sl and TP.  That is why I asked this question.  Because some EA's have this functionality.   

 
dustovshio #:

Your if your EA has an opposite signal it can close orders.  If at boost can theoretically reduce false signals.  They if the opposite signal closes orders, then catboost would reduce the false orders from closing and the result would be you leave orders open longer and greater profit is achieved.  For example. You place an order when your MA crosses.  your stoploss is 50 pips and TP is 50. However, there MA crosses back over before you it your SL or TP, and your EA is programed to close the order it his happens: this is referred to as close (or reduce, or reverse) on the opposite signal.  Now if that signal was a false alarm, then you closes your profit too early, when it could have gone up to your TP instead.  So catboost could have filtered a certain percentage of those false signals? This is my question.  Not all EA's close positions on the opposite signal. Many just have a fixed Sl and TP.  That is why I asked this question.  Because some EA's have this functionality.   

I understood what the conversation was about.

Programmatically, it's easy to implement, but it will be a game with randomness. The fact is that the Recall index in the models is quite low, i.e. the model does not recognize more than 10% of all events, which means that the opposite position is often not open due to an unidentified pattern. This is, among other things, related to predictors. The article shows the algorithm for implementing CatBoost models. It is necessary to strengthen the model with predictors, then your proposed approach will be more justified.