Suppose we have know the statistics of items.
For Each item we can define 4 values:

A is the number of winning games with the item
B is the number of losing games with the item
C is the number of winning games without the item
D is the number of losing games without the item

We know that:
A/(A+B) is the Win Rate of the item, and we call it Wi.
(A+B)/(A+B+C+D) is the Pick Rate of the item, and we call it Pi.
(A+C)/(A+B+C+D) is the Generic Win Rate, and we call it Wg.

Consider this Ratio: A/(A+C)
The number of winning games with the items among all winning games.
This ratio tells us how much the item is "necessary" to reach the victory. So we call it the "Necessity Rate" or more simply "Need Rate" Ni.

Consider the Wi
The number of Winning games with the items among all games with the items.
This ratio tells us how much the item is "sufficient" to reach the victory.

So, we can say that an item is the best if is "necessary and sufficient", by making the comparison of Wi*Ni of all items and choosing the greatest ones.

But what is Ni in terms of Pick and Win Rates?

Ni=A/(A+C)= [A/(A+B)*(A+B)/(A+B+C+D)*(A+B+C+D)/(A+C)]= (Wi*Pi)/Wg


So an item i1 is better then an item i2 if

Wi1*Ni1>Wi2*Ni2

or in other words[/h2][/h4]

(Wi1)^2 * Pi1 > (Wi)^2 * Pi2

(Wg is canceled because it's the same with all items).

The Final Result of this discussion is:

"The square of Win Rate times the Pick Rate is the quantity to compare among items to choose the best build"