I don't like the standard +=, =, =+ and so forth framework for communicating positional evaluation, because it doesn't do enough to convey the winning chances of the two players.  It is not a one-dimensional, but a two dimensional problem.

Here is what I propose, in perfect confidence that it will instantly be adopted by the chess world:

Certain:

WW   White wins with best play
==   The game is drawn with best play
BB   Black wins with best play

Uncertain, Undynamic (scant winning chances for at least one side):

W    Good winning chances for White, scant ones for Black
w    Some winning chances for White, scant ones for Black
=w   Scant winning chances on either side, White's play is easier
=    Scant winning chances on either side
=b   Scant winning chances on either side, Black's play is easier
b    Some winning chances for Black, scant ones for White
B    Good winning chances for Black, scant ones for White

Uncertain Dynamic (at least some winning chances for each side):

Wb   Good winning chances for White, some for Black
zw   Some winning chances for each player, White's play is easier
z    Some winning chances for each player
zb   Some winning chances for each player, Black's play is easier
wB   Good winning chances for Black, some for White

Zw   Good winning chances for each player, White's play is easier
Z    Good winning chances for each player
Zb   Good winning chances for each player, Black's play is easier


This is a total of 18 possible marks. Note that "z" can be thought of as shorthand for "wb," "Z" for "WB."  Note also that given certainty, everything is WW, == or BB; all else presupposes uncertainty, which is why judgment is called for.  

So, how would I evaluate the initial position?  zw.

Fair enough.  So, propose a system of your own.

Just look how these people mock my genius.

But to continue, the two-dimensional evaluation system facilitates the application of alternative criteria to the selection of moves (good players do this already; it's only that the evaluation system calls attention to it).  Here are some possible criteria for White and the ranking of evaluations that follows from them (rankings are in order of decreasing desirability; also since all rankings start with WW, W and end with B, BB, I omit these).

1.  Maximize relative chances, maximize own chances among ties.  This gives: Wb, w, Zw, zw, =w, Z, z, =, ==, Zb, zb, =b, wB, b.

2.  Maximize relative chances, minimize opponent's chances (chance of defeat) among ties, maximize own chances among further ties (this last necessary to choose between Zb, zb and =b): w, Wb, =w, zw, Zw, ==, =, z, Z, Zb, zb, =b, b, wB.

3.  Maximize own chances, maximize relative chances among ties (almost like, White must win): Wb, Zw, Z, Zb, w, zw, z, zb, wB, =w, =, ==, =b, b.  But if White must win, then "==" should be shifted to the bottom.

4.  Minimize opponent's chances, maximize relative chances among ties (White must draw): ==, w, =w, =, =b, Wb, zw, z, zb, b, Zw, Z, Zb, wB.

Clearly these different criteria would give rise to very different move selections.  This already happens in practice, it's only the the marking system illustrates how.  

A single tree of variations marked with this method and "backsolved" according to some criterion would readily yield the optimal move selection under that criterion, but only if one assumed the criterion according to which each side is selecting its moves!  A standard assumption would be that the opponent is merely trying to thwart one's goal, but that isn't necessarily the case.  If one can assume that one's opponent must play for a win, while one's goal is to maximize one's relative chances, the opportunity may arise to steer the game (as White) into Wb territory since the opponent will prefer this even to =b.

My point is that to mark a given position in a tree of variations with fixed terminal evaluations, one needs to assume the criteria under which either side will select its move.  The mark assigned to any given position will depend on the two criteria assumed.  

This aspect of move selection is not revealed, but rather is concealed, by the standard, one-dimensional system of marking positions.  Under a one-dimensional system, for a given set of terminal marks, there is always a set of equally preferred, objectively optimal moves at the root position.

I will conclude by remarking that verbal annotations, while fully sufficient for conveying a message like Zb, do not suffice for ranking positions and backsolving trees of variations, or if they do suffice, there must be a very laborious process of translation and interpretation, which could have been avoided if the terminal positions had been marked according to the proposed method.

Thus we need some further modifications to arrive at an algorithm that will lead to more meaningful probabilistic determinations appropriate for accurate back-solving capabilities. 

For example, I would propose the following:

WbPTdW   Good winning chances for White, some for Black, where White is a predominately positional style player, Black is a predominately tactical player, White will be happy to draw, while Black needs to win.

I suggest you likewise expand on each of your evaluation acronyms to take into account these additional dimensions. 

Just look how these people mock my genius.

But to continue, the two-dimensional evaluation system facilitates the application of alternative criteria to the selection of moves (good players do this already; it's only that the evaluation system calls attention to it). 

Actually, I like this idea a lot. Current symbols don`t really tell much. But it is important to keep the symbols as simple as possible, lest you have to look up in a dictionary everytime you want to decipher the notation. For instance we could just keep 

+-:White wins with best play
=:   The game is drawn with best play
-+:   Black wins with best play