MadBill.... you're correct. I believe we went through  total test series that .5-day. At  sequential cylinder firings per series to analyze. My brain was fried by noon! Putz.... the  'mis-fires' average out of  sequential firings was the best-case. I was using AutoLite/MotorCraft plugs at the time and the best case was for a projected-nose, J-gap plug. We ran  series each with non-projected nose and projected nose with the ground electrode covering the center electrode. Those  series revealed a higher % cyclic variation. Up to about 13% with the non-projected nose. We didn't try any 'indexing' tests at all. My dyno time and $$$ ended after .5-day. And yes...... the best-case test series showed  additional ft/lb of torque and  HP on the high end. This on a 2.0L ~232HP/171-torque assembly. Even with the best-case plug and a slam-bang(I believe!) crank-triggered ignition, the engine still mis-fired during the last  test series. With what the SAE paper had to offer, and MadBill's comment, I'm currently convinced cylinder/chamber swirl and tumble is a majority definer/contributor to cycle-cycle variations. Or lack thereof. Mixture motion near/around the developing flame kernal. And in the case of the Honda 4-valve used in the SAE paper, plug index/orientation was also a moderate 'fix' for cycle-cycle firing variations. I'm also convinced(no engineering data!) that there are induction/head/piston combinations that are so terrible with regard to swirl and tumble, particularly tumble, that no plug, orientation, or ignition system on the planet could completely prevent some amount of cycle-cycle variations in cylinder firing. Although you gents with 'wet-flow' hardware are getting us much closer, without mega-$$$ test hardware, than we were even a very few years ago. Even if it is 'steady-state'.