Method 1 is close and Method 2 is even more off. There are a few mistakes:
The dB is a relative value, it is always calculated from a ratio of something. It is up to you to know what the reference is. You cannot convert direct values to and from dB and apply them correctly without keeping track of how those values were calculated in the first place.
10log(165) - 10log(1)
is wrong; 1W is reflected; remember, subtraction of logarithm yields the same as division of numbers the logarithm is taken from.
By the way, log(1) is always 0, so this should be a giveaway that this operation can't be right.
So, 1.3 dB = 1.3489 W
no. Definitily not. 1.3 dB is a unitless factor.
You need to refresh what a decibel is,...
If your error vector magnitude is below the decision threshold between adjacent symbols then there would be no impact on the BER. Thus there will be a point where you can continue to reduce the error vector while not see a change in BER.
Ultimately our objective is to minimize the error vector regardless, to maximize performance in lower SNR conditions. Also ...
Find a good pretrained audio classification model, collect some sample data, fine tune the model for your task, use it for inference.
Check teachable machine, check this video to see it working. It will make the task to look a lot easier than it could be.