It seems that they (TechPB) are still restoring the database and, I've seen a bit more of the discussion (and more importantly the data) but, there's only three pages worth.
Here's my current thoughts: compare only the data for each individual due to size, speed and arm movement differences between individuals. Unfortunately, Bryce has the largest sample of runs but, not enough at each BPS to get a good statistical sample. They would probably admit as much. But, throughout the data, they prove that it's possible, even at 15BPS to run through an established, non-tracking lane. But, there's not enough data to support probability claims.
As far as accuracy goes, Vortex shedding has been shown to jack up the trajectory of a perfectly smooth sphere- see any number of aerodynamics studies. A paintball with it's randomly oriented seam, and slightly ellipsoidal shape (again in random orientations) will certainly not do any better, especially with random spin (in magnitude and orientation). I believe that Cockerpunk would agree with me on this. In any case, I don't believe they are done with 'accuracy' testing. If the forum date info is correct, they did another bore fit test in late September, and that thread ends with several folks and them considering further testing to nail down some theories.
In regards to their accuracy statements- to be clear, their vector and Standard deviation values reflect projectile spread and not probability of impact which, could be influenced by the size of the target, how effective are the aiming method and devices are, the player's abilities, etc. The use of a vector is due to the fact that there has been negligable elliptical bias seen in the data that appeared across any single test session. I can't be bothered to crunch the data now but, I suspect that if one were to examine this data
they wouldn't even be able to correlate shot velocities with impact deviation in the Y-axis.
I believe that you can apply the vector to determine your probability of hitting someone with a single shot. For example, consider the CP .685 14" barrel data. Let's say you sight in on those 20 rounds so that the center of the group is centered on your reticle. In this case, it would you would sight in at the -3.14,-4.36 (x,y) position on the board. While keeping the barrel locked in, you fire another set of shots. Assuming that 20shots is enough of a sample size, 68% of your following shots should land within a radius of 14.53" If you double the radius, it should contain 95% of the shots. Therefore, if you wanted to be 95% sure of hitting your target with a single shot, don't aim at anything smaller than 58.12" tall/wide when 100' away.