In my work I have found two forms of mistakes. There can be major ones where a formula is being calculated entirely wrong due to a human error and others where two data surces will vary very slightly in their numbers on the same thing.
What I found was most useful correcting a back test was replace the data in question close to and around trade signals to see if it afects the general system function.
In my own case, a poster named vbontchev was very upset that my vix numbers were .001 differnt from his. After four hours work painstakingly correcting this, it did change one thing about the signals from my program.
======================================== Possibly 20 more mistakes that I'm aware of. Since receiving the spreadsheet from tclark I've been focused on NASI #'s, formulas and adv/dec #'s. ========================================
So what errors have you found?
I am not familiar with the data sources used. I would guess that the NASI data came from the same source as these:
I have spot checked the data on the ss with this source and it all appears to be correct.
The functions used come from TA-lib.org and I am assuming they are reasonably correct though they don't always match Stockcharts.com values as mentioned in previous posts, I believe.
The exact NASI calculation used is shown on the NASI page. I am not familiar with the particular NASI calc you mentioned previously.
What data sources/functions would you suggest we use?
Could you please list out the 20 possible problems. If we don't stay together on these corrections/refinements, we'll end up having multiple spreadsheets floating around.
Market Data 
Markets 
AI
