Some ballpark information from racing manifold builder I found some time ago.http://www.swartzracingmanifolds.com/tech/index.htm
I quantify runner taper by comparing the cross-sectional area of the intake port (at the gasket) versus the area at the plenum opening of the runner (not including the entry fillet). This is expressed as a percentage of increase in area. I calculate the area at the intake gasket, and then I add the percentage required by the application to determine the runner entry area. Since most applications have round plenum openings, I simply convert the opening area back to the diameter of the required circle.
Obviously, the key is to know what percentage is needed for your application. Several factors seem to be important to determine this number. I can’t quantify the impact of these factors, as every engine combination is different. Camshaft events alone can severely affect how taper is tolerated. I can give you the following relative guidelines to determine where to start:
•Most drag racing manifolds require between 25% and 45% increase in area. There are applications that go beyond these limits, but this is rare.
•Engines that are built on production platforms tend to like about 30% increase in area. I’m referring to the typical bore/stroke/rod ratios derived from OEM engines. As bore/stroke ratios approach 1:1, the increase in area is likely to be 25% or less.
•Race engine platforms with large bores and short strokes like higher percentages – 35-45%.
•Short deck engines (and resulting rod/stroke ratios) tend to want more taper. The piston motion of a short rod combination will create adequate port velocity in large runners.
•Stick-shift applications tolerate more taper than automatics.
•Larger displacement engines like more taper than small engines of the same bore/stroke/rod proportions.
These guidelines should help you narrow down your taper requirements within a range of a few percent. The combinations are so varied, even within the same class of racing, that set numbers and fixed rules are difficult to derive. My goal is to minimize your testing iterations required to find the best performance.