Here's some code for creating a layer (named "lines_layer") with half of the rows randomly selected, which will work with ArcGIS 10.1+:
import arcpy
import random
fc = r'C:\path\to\features.shp' # Path to feature class with attribute table to modify
object_ids = [r[0] for r in arcpy.da.SearchCursor(fc, ['OID@'])]
sample_size = len(object_ids) / 2 # To select more or fewer than half, change denominator
random_ids = random.sample(object_ids, sample_size)
oid_field = arcpy.Describe(fc).OIDFieldName
selection_query = '"{0}" IN ({1})'.format(oid_field, ','.join(random_ids))
arcpy.MakeFeatureLayer_management(fc, "fc_layer", selection_query)
Just change the fc definition to point to the full path of the feature class or shapefile whose fields you want to calculate. (A layer name would work also, if you've already made a feature layer from it. In that case, half of the features in the layer would be selected.)
In your question, it sounds like selecting every other line may be good enough, in which case I would direct you to the query used in my answer to this question.