Add or reduce timing?

Here's the result of the changes applied to my base timing table:


As you can see, my "Feedback Knock Correction* (degrees)" has actually increased i.e. it has got worsed :(

I did a search on NASIOC and someone said that increasing timing could actually help:
"FYI, on my 02 WRX, I have found that I had to increase the timing quite a bit below 2000 rpm all the way out to 1.2 load column. I think what is happening is the mixture does not have enough time to burn, so when it exits, it is still burning which ignites the incoming right away. I can drive in 5th gear at 35 mph with no bumping, shaking, etc. I can also do a 5th gear pull from 1200rpm and have it nice and smooth. I'm still working on it though, but the result has been increase the timing.

what helped me figure it out was making a 3D plot of g/min divided by rpm(based on your map table). it bascially gives you the g/rev, then you can divide(or multiply) it by your timing value in the map. it will show you spots that don't quite line up with the overall trend, the problem is you don't know what the trend is, or where the ends shoudl line up, so you have to keep that in mind. I think you will find your low rpm, mid load range is off.

my thinking is that if you keep the rpm consant, and increase the load...then the charge will need less time to burn, so lower timing is needed. if you keep the load constant, and increase the rpm, it needs more time for a complete burn...thisonly makes sense for closed loop where the AFR is constant. I have not done the physics research to see if this increase is linear or not, but I assumed it was and have had good results. Once I get it lined out, I will start to increase timing and see what happens. I also zeroed my CL/OL delay and made everything below 2000 rpm run in closed loop...

one more thing, take your log and graph only those load/rpm points where you have FBKC or FLKC. I make an X/Y scatter plot with rpm in the y axis, load in the x axis, and then scale the axis to match my timing table values. I did that to your logs you posted...and you can see exactly where you need to focus."

Here's the full thread:
http://forums.nasioc.com/forums/showthread.php?t=1498977

Ok, I'm going to give that a shot. Here's my change (20080530_OL9b_ravenhil_cracked-masv1b_basetimingonly.hex):