Michael,
Whoa there cowboy! First off lets go easy on the mathematically challenged. I think the reason most people don’t get the underlying math is that in the real world, (home, jobs, daily life), the vast majority of people don’t even use it. Thank the education system, calculators and the industrialization for that. Henry Ford doesn’t need mathematicians to tighten down the seat, Amazon isn’t hiring statisticians to deliver packages. Common Core is designed to provide workers, not thinkers. On the other hand, do you know how to make lavened bread from soil and seed? Properly temper a knife you hammered yourself in a homemade coal forge? Unlikely because they sell Wonder Bread and knives so inexpensively. : ) Nuff said.

On to the real discussion. You stated:
‘The only way to attain believable results is to use decimal multipliers, like x0.8’

Hum..???… I think your right.
If I remember correctly CoC 7ed and Mythras attempts to address the -20% ‘high meaningless, low skill massive’ by tests which are a fraction or multiple of your base skill. Like hard being 1/2 skill so a 50% skill becomes 25% and a 20% skill becomes 10%. What I don’t like how it penalizes high skill characters much more than low skill characters. A master with 100% who would normally succeed 9.5 out of 10, (96-00 always fails) now fails 5 out of 10 times while a stooge, (20%), who normally fails 8 out of 10 times now fails 9 out of 10 times.

Your x0.8 solves this but becomes math intensive at the table. “You have 77% skill with a x0.8 mod for range, x0.5 mod for cover and x1.5 mod for the ‘seeker’ spell on the arrow.
I am not good enough at math to think of an easy way to overcome these multiple mods in play. Any ideas?

So the bell curve of 3d6 does seem to work better for mods than a % system. What you lose is granularity and steady skill progression. Each point of skill gained is now referenced to your current skill. 1 point will have a value 0.46 to 12.5. A single ‘bit’ of information learned by someone in the middle, (such as the names of a couple demons for a sorcery skill), might increase his skill chance by 0.46%, (2 skill becoming 3 skill using 3d6), or 12.5%, (10 skill becoming 11 skill using 3d6). It is also far less obvious that if I have a 12 skill on 3d6 that it equals a 74% chance than 74 Skill equaling 74% using a percentile roll. I believe we tend to think in ‘slices of pie’ more than bell curves.

Perhaps skills increasing as percentages yet rolled on 3d6 to the nearest percentage would work better when modifiers are applied? Either that or still rolled on percentile dies but your fractional modifiers would be charted out for ease of use. That’s gonna be a big chart to cover all the 100 skill levels and multiple mods.

I have also though about things like Stat Tests being based upon the highest Stat in the contest rather than a static 3d6 under roll or a Resistance Table, (Basic Roleplaying), where each point of difference equals a 5% modifier.
1-5 = 1d6, 6-11 = 2d6, 12-17 = 3d6, 18-23 = 4d6, etc; Where the the highest Str is always at least 1 point above the maximum die roll. All 1’s is always a success and all 6’s always a failure. Thus when when a giant with Str 30 is arm wrestling a Hobbit with Str 8 they both roll 6d6. Winner is the one who rolls highest but also under their Str. Its not so much for the contests of 8 vs 30 as it is for 25 vs 30. A mechanism where each point is a 5% modifier means a 6 vs 12 is the same as a 12 vs 18, 20 vs 26 or a 100 vs 106. This is fine IF each point is significantly higher than the previous in terms of its value where a 106 Str carries twice as much weight as a 100 Str, (being equivalent to 6 vs 12 or 12 vs 18), but fails if the points are of equal value. Also, the mechanism of dding the result to a die roll such as 1d20 + Str where the highest result wins is fine when a 6 vs 12 but not when a 20 vs 39 or 100 vs 106.
This came up when I was designing a d100 campaign which revolved around warring giants of various clans and types and realized the mechanism for resolving Stat vs Stat battles was not very good.
Any thoughts on this?

Thanks for your insights and input!
Mike