Why do the DLPredictor indicator values change depending on the strategy start date? The strategy signals and all strategy parameters change accordingly.
For example, if the strategy backtest starts in 2001, the last indicator value is -0.27.
But if I change the backtest start to 2003, the last indicator value will be 0.65 and all strategy characteristics change.
There are no indicators with parameters greater than 200 bars at the DLPredictor indicator input.
For example, if the strategy backtest starts in 2001, the last indicator value is -0.27.
But if I change the backtest start to 2003, the last indicator value will be 0.65 and all strategy characteristics change.
There are no indicators with parameters greater than 200 bars at the DLPredictor indicator input.
Rename
I tried different options for input layers of the neural network, including the simplest option when there are only 2 layers at the input - RSI and SMA indicators, the problem persists.
I usually set up a strategy for a long period, and in the strategy monitor I set a shorter period (2-3 years) so that all strategies in the strategy monitor are processed faster. But it turns out that all strategies do not work as they did when the strategy parameters were initially set up.
I usually set up a strategy for a long period, and in the strategy monitor I set a shorter period (2-3 years) so that all strategies in the strategy monitor are processed faster. But it turns out that all strategies do not work as they did when the strategy parameters were initially set up.
I'd guess it's because RSI is not a completely stable indicator, and while it converges to nearly identical values, the values are not identical when the start date changes. In contrast, here's a DL indicator based on ConsecUp/ConsecDown which are stable indicators. Even though one chart begins in 2001 and another in 2003, the indicators have exactly the same end value.
I have RSI indicator in many strategies (not neuro), but their profitability parameters do not change depending on the start of the backtest. This happens only with neuro strategies.
And if this is so, then a list of indicators is needed that cannot be used in the DL extension due to their "instability"
Re: Post #4. You will find it straight in the Help: ask Co-Pilot.
The Simple Moving Average (SMA) is generally stable as it is not calculated using its own previous values.
I made the DLPredictor indicator which only has SMA indicators of different lengths at the input. I run it from 2001 and from 2005 and again I get different results.
I made the DLPredictor indicator which only has SMA indicators of different lengths at the input. I run it from 2001 and from 2005 and again I get different results.
Your Response
Post
Edit Post
Login is required