I created a spreadsheet for this project and it actually reveals very little by itself except stat disparities (and that is subjective itself since without an empirical guide to assigning weights to power and toughness).
After you enter casting cost, power/toughness, and the list of keyword abilities, I created an X-factor listing which was the evaluation of the rest of the cards abilities. It turns out that the ranking of the cards hinges almost entirely on that "X-factor" entry.
For instance, almost every white 1 drop is a 1/1. The difference between Doomed Traveler and Goldmeadow Harrier comes down to how you weight their abilities. This is assuming there is a fixed weight table for keywords (ex: Akrasan Squire).
Also, I devised a minor formula for making colored mana cost drawbacks: (colored mana-1)/(cmc-1). For cmc=1 I set this value to 0.
the method is decidedly not perfect especially because it evaluates cc, ccc, cccc, ccccc, all as -1 but its better than a flat -1 across the board without regard to whether cmc=1 or cmc=10
having random modifier of .33 or .67 will just make the math more complicate, without getting more exact i guess. Although you use it in an excel table. Also you will notice that those modifier are even worse than a flat -1. CCC1 is uncastable but has less penalty? There are none at common anyway. And those numbers are out of the blue, since ive already shown that double mana relates to stats 1:1.
I also still wonder how actual game experience differ so much, since i cut almost all double costed cards from my cubes because theyre so inconsistent (up to CC5)