How to choose the best filter for White Noise from the Acceleration Az Signal of a MPU6050 connected to Arduino?

I have to analyze my Az signal to identify road anomalies, from my MPU6050 connected to Arduino Uno. I have a Civil Engineer background so i am noob on Signal Theory and Filters

On Matlab i Implemented two filter:

1. Savitzky-Golay
2. Butterworth

I conducted a static test, recording my signal without perturbation on a table. I obtained the following results.

This is the signal after Butterworth with this cutoff value [b,a]=butter(2, 20/25); (i can't understand the intial peaks)

This is the signal after Butterworth if i remove the first peaks

vbuttercleaned=vbutter(100:end,:); 
>> t_butter=time_resampledf1(100:end,:);

This is the signal after Savitzky-Golay Smothing filter with this parameters vsgolay = sgolayfilt(vq1,2,27);

1. How can i choose the best filter for my purpose?
2. Why I Have that peaks with the Butterworth Filter?

I thought to implement a pilot test and analize false negatives ad false positives. The standard deviation before/after filtering is a significative measure?

I have a civil engineer background, so is my first time with signal analysis, never studied Signal Theory before.