this is my search:
| makeresults count=2 | 
eval start=relative_time(now(),"@d") | 
eval start_string=strftime(start,"%Y-%m-%d") | 
eval end=strptime("1 jan 2017","%d %b %Y") | 
eval end_string=strftime(end,"%Y-%m-%d") | 
eval start_minus_end=end-start/1400
give s me the following:
_time   end end_string  start   start_minus_end start_string
1   2016-04-20 10:13:02 1483182000.000000   2017-01-01  1461067200.000000   1482138380.571429   2016-04-20
2   2016-04-20 10:13:02 1483182000.000000   2017-01-01  1461067200.000000   1482138380.571429   2016-04-20
How do I work out how many days are between start_string and end_string? 
I have tried to do this with eval start_minus_end=end-start/1400 but i am not sure what value start_string is? is it milliseconds?
As an aside I would then like to store this value in a variable that I can somehow use in  the predict function and assign it to future_timespan  e.g. predict  kpi1 as kpi1 future_timespan=<VARIABLE> holdback=1
 
					
				
		
Splunk also has a tostring function that allows you to, among other things, convert seconds to a duration. The result is in the format of days+HH:MM:SS
| eval duration = tostring(end - start,"duration")
So in your case the result for today is 257+01:00:00.000000
 
					
				
		
 
		
		
		
		
		
	
			
		
		
			
					
		