Introduce Signal Strength Treshold


An important feature can allow the evaluation of the strength of the trading signal. This is accomplished by looking at the unthresholded output generated by the NN. For example, a very strong buy signal corresponds to an output close to 0.5% or 1%. We study the possibility of validating a trading signal only when the output of the NN exceeds a specified non-zero threshold instead of just applying a sign-function. This threshold is a parameter that even can be adaptive in theory.
But for simplicity choose this threshold to be 0.2%.

1 dimension input case: only SMA(180) is used. nNeurons = 5;

  • threshold = 0: winLoseRatios Arithmetic Mean: 53.23%, stdev: 4.30%, avgBarGainPercent mean: 0.15%, stdev: 0.28%, projCAGR: 7.94%
  • threshold = 0.2%: winLoseRatios Arithmetic Mean: 53.76%, stdev: 8.19%, avgBarGainPercent mean: 0.09%, stdev: 0.22%, projCAGR: 4.97%

2 dimension input case: SMA(180)/SMA(20) are used. nNeurons = 5;

  • threshold = 0: winLoseRatios Arithmetic Mean: 53.68%, stdev: 4.16%, avgBarGainPercent mean: 0.16%, stdev: 0.27%, projCAGR: 8.56%
  • threshold = 0.2%:winLoseRatios Arithmetic Mean: 54.92%, stdev: 10.03%, avgBarGainPercent mean: 0.09%, stdev: 0.18%, projCAGR: 4.88%

The conclusion is that it improved the W% WinLose percents; so it was more accurate, but it missed many trades (about 50% of the trades were not played), so the CAGR is reduced to half.
I cannot see the performance chart, but there should be less drawdown with the new system, therefore the Sharpe ratio may be increased.

I leave this code part in the source code, with the Signal Strength Threshold = 0 as a default, but we will study its effect again later.


No Responses Yet to “Introduce Signal Strength Treshold”

  1. Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: