How to efficiently generate a percentage difference series from conditionally selected values?

I'm working on a particular problem which I managed to solve in a very inefficient way and I would like to see if there a more efficient way to do it. I have a Dataframe that has its first column filled with positive values and then almost 69 columns filled with either 1, 0 or -1 that represent a particular condition for a certain property. I divided these columns into 4 groups based on their position and added a column in the last position (69) that has all null values:

Group A = (69, 5, 6, 9, 10, 12, 15, 19, 20, 27, 43, 53, 54, 55)
Group B = (69, 6, 10, 15, 29, 31, 34, 36, 48, 49, 58, 59, 60, 62, 64, 68)
Group C = (69, 7, 13, 16, 17, 22, 24, 26, 28, 32, 33, 38, 40, 42, 44, 46, 51, 56, 66)
Group D = (69, 8, 11, 14, 18, 21, 23, 25, 30, 35, 37, 39, 41, 45, 47, 50, 52, 57, 61, 63, 65, 67)

What I have to do is to randomly pick 1 column from each group and check if either all the values are 1 or -1 while Null values provided by column 69 counts as wildcars, so if i have 1,null,1,null counts as 1,1,1,1 and viceversa.

I have to generate a series with the percentage differences between values of the first column of the dataframe. The first value will be the one in the first row that has all -1/null, the second starting from this last row will be the first that has all 1/null this would be the first difference, from this row will go down the whole dataframe selecting the first -1/null and the subsequent 1/null to create a second, third, fourth, etc percentage differences adding them to a series.
This is how this particular array will look like:

Values     Column 6    Column 69    Column 28    Column 61     Selected             Perc.Diff 
  43          1           Null          1             1           43 (1st 1/null)       Nan
  90         -1           Null         -1            -1           90 (1st -1/null)      Nan
  20          1           Null          1             0           Nan                   Nan
  40         -1           Null         -1            -1           40 (2nd -1/null)      Nan 
  30          1           Null          1            -1           Nan                   Nan
  95          1           Null          1             1           95 (1st 1/Null)      0.055 (90->95)
   8          1           Null          1             1           8  (2nd 1/Null)       Nan
  52          1           Null          0             1           Nan                   Nan
  63          1           Null          1             1           63 (3rd 1/Null)       Nan 
  73         -1           Null         -1            -1           73 (1st -1/null)      Nan 
  59          1           Null          1             1           59 (1st 1/Null)     -0.274 (73->59)
  19          1           Null          0             1           Nan                   Nan
  21         -1           Null         -1            -1           21 (1st -1/null)      Nan

The last 2 columns doesn't need to be in the dataframe since i only need a series with the diff values that doesn't need to have the same length as the dataframe. I would like to see if there is a more efficient way to do this. Here is my solution:

a = shuffle([69, 5, 6, 9, 10, 12, 15, 19, 20, 27, 43, 53, 54, 55]) 
b = shuffle([69, 5, 6, 10, 15, 29, 31, 34, 36, 48, 49, 58, 59, 60, 62, 64, 68]) 
c = shuffle([69, 7, 13, 16, 17, 22, 24, 26, 28, 32, 33, 38, 40, 42, 44, 46, 51, 56, 66])
d = shuffle([69, 8, 11, 14, 18, 21, 23, 25, 30, 35, 37, 39, 41, 45, 47, 50, 52, 57, 61, 63, 65, 67])

Generate list with 4 random column indexes:

ind_list = [a[0], b[0], c[0], d[0]] 

Create series based on selected values:

lim = ind_list.count(69)
df.loc[(df.iloc[:, ind_list[0]] + df.iloc[:, ind_list[1]] + df.iloc[:, ind_list[2]] + df.iloc[:, 
ind_list[3]] == len(ind_list) - lim), 'Selected'] = df.Values

df.loc[(df.iloc[:, ind_list[0]] + df.iloc[:, ind_list[1]] + df.iloc[:, ind_list[2]] + df.iloc[:, 
ind_list[3]] == -len(ind_list) + lim), 'Selected'] = -df.Values

Filter values that are not "Firsts":

l1 = 0
df1 = df[df['Selected'] != 0]
df1.loc[(df1.Selected.shift(1).apply(np.sign)== df1.Selected.apply(np.sign)), 
'Selected'] = 0
df2 = df1[df1['Selected'] != 0]

Removing first value if it has been selected with the 1/null condition since first element has to be -1/null and the last if it is -1/null. Column 70 is the 'Selected' Column.

if len(df2) > 0:
   if df2.iloc[0, 70] > 0: 
         l1 = 1
else:
   continue

if df2.iloc[-1, 70] < 0:
     perc = df2.iloc[l1:-1, 70].abs()
else:
     perc = df2.iloc[l1:, 70].abs()

Calculate percentage differences: perc_diff = perc.pct_change()[1::2]

The problem with this method is that if/else block that I feel could be avoided and also a SettingWithCopyWarning for this line of code:

df1.loc[(df1.Selected.shift(1).apply(np.sign) == df1.Selected.apply(np.sign)), 'Selected'] = 0

That I don't know how to avoid. Thanks for all the help you can give me, since this script is part of a loop that has like thousands of iterations any ms that I can save counts.