ta0 made some excellent points regarding top design. Definitely read the paper he recommended: Bächer et al., 2014, Spin-It: Optimizing Moment of Inertia for Spinnable Objects. (Free PDF
here.) Many important practical concepts for the top designer, including "spinnability" -- a big part of play value.
Since everything affects everything else in a top, design comes down to playing the many trade-offs just so. I usually start by setting some performance and esthetic goals to guide my choices. These common goals often conflict:
o Play value (almost always my primary goal)
o No wobble or hop (usually 2nd priority)
o Easy launches by hand without scraping, or high-speed launches with a dedicated starter
o Some cool visual effect at speed
o A particular spin-down behavior -- e.g., sleep, steady precession, self-righting, or fast travel
o A particular look at rest
o Performance in battle
o Longest possible spin time given higher priorities
You'll have to experiment to get a feel for the trade-offs involved. Many common goals carry a spin-time penalty.
Yes, a disk with a light stem is almost the perfect top with respect to weight distribution. Even better, if you replace the disk with a ring (and light spokes).
"Perfect" and "better" for spin time, but not necessarily for other goals. Below are 3 LEGO finger tops with peripheral mass concentrations and the same small ball tip.
Left: Mass = 48.6 g, max radius = 34 mm, CM height = 25 mm, sleeping spin time by hand = 152 s. Since the black rubber tire is a good bit denser than the ABS plastic, most of the top's mass resides near max radius. And the center of mass (CM) is quite low. Combine this very favorable "down-and-out" mass distribution with good aerodynamics, and you get over 2.5 minutes of spin time.
Middle: Mass = 40.5 g, max radius = 34 mm, CM height = 33 mm, sleeping spin time by hand = 68 s. Max radius same as above, and the aerodynamics are comparable. But the mass is 17% smaller, and the mass distribution isn't as down-and-out. Result: 55% shorter spin time, but I really like the way this one looks -- both at rest and at speed.
Right: Mass = 54.8 g, max radius = 64 mm, CM height = 30 mm, sleeping spin time by hand = 14 s. The mass distribution has 6-fold rather than circular rotational symmetry, but it's still fairly peripheral. Yet spin time totally tanked -- mainly due to the very dirty aerodynamics.
As you can see, spin time correlates rather poorly with total mass, max radius, and CM height. Instead, it's mostly about radial mass distribution and aerodynamics.
But spin time isn't everything. The top on the left may spin a long, long time for a LEGO top, but it's kinda boring to watch. The top on the right, on the other hand, spins only 9% as long, but the lime and orange of the spokes mix to match the gold of the the central ornament at speed. I really like color-mixing effects.
But it actually does not have to be a circle or even have any radial symmetry. Any flat shape will be stable if spun around a perpendicular stem going through its center of mass.
That's true only to a point. Below are 3 perfectly balanced simple LEGO test tops differing only in max radius, rotor aspect ratio, or both. Those at far left and far right have 1:1 aspect ratios. And both are quite stable at finger speeds despite the four-fold difference in max radius. But it takes a very lucky twirl to get the one in the center (2:1 aspect ratio) to stay up at all. Yet any aspect ratio between 1:1 and 2:1 spins quite nicely.
Except for color, the 2 assembled tops below use the same parts (here laid out between them) and therefore have the same mass. But the one on the left (1:1 aspect ratio) stays up easily, while the one on the right (4:1 aspect ratio) crashes to the ground right away every time.