Saturday, December 2, 2023 5:17:28 PM

Number of trades

one year ago
#35 Quote
I'd like to ask if there's a logical approach or procedure in order to decide how many trades per day I expect or want. After several model builds I've obtained from 0 trades in 1 year to 3000 trades per day; I've noticed that varying the pre-signal window and target window somehow (in a way that I don't understand) directly influence the end results but these results are unpredictable.

Could You please give me some kind of procedure or guidance?
one year ago
#36 Quote
Hi Alberto,

Thank you for posting. Here is a link to how the machine learning model is created along with more detailed information, but there are a few parameters which will affect the number of trades.

Ticks Up To Target/Ticks Down To Target - These parameters determine how many ticks up or down the price of the instrument needs to go in order to reach the profit target when creating a machine learning model. As an example, if you are trading ES with Bars To Target set at 5 bars and the Ticks Up To Target is only 1 tick, then the model will be created using data that only went up one tick. In this case you may have many data sets that are used when creating the model and the model may create many trades. If you have Ticks Up To Target set at 20 ticks then there may not be as many trades.

Bars To Target - This parameter will determine how many bars in advance the model should look for when reaching a profit target. The price of the instrument needs to move up Ticks Up To Target or down Ticks Down To Target to reach the profit target. If the Bars To Target is set at a large number then it may be easier to reach the profit target than a smaller number of bars.

Pre-Signal Window - This will determine how many bars of data the model will use when creating the model, along with how many bars of data the model will use when predicting a trade signal. The larger the Pre-Signal Window the more the model will try to fit a particular data set.

Data Series - The Data Series that you are using will greatly influence the number of trades. As an example, if you are using 1 minute bars versus 1 hour bars.

Indicator Data - This is the data that is added via AddDSTIndicator. When doing a prediction, each of the indicator parameters will be sent to the machine learning model to determine if we have a trade signal. The greater the number of parameters the more that the indicator data will need to fit the model.

You will need to experiment a little to find a sweet spot for the particular instrument you are trading. Trying different values for ticks up to target, ticks down to target, bars to target, data series time frames along with indicator data may help.
one year ago
#37 Quote
ok,now it's more clear! thanks for the explanation.
one year ago
#38 Quote
Glad to help!