Just a counterpoint, you've stated your method, now how good does it actually work out for you? Or otherwise stated more bluntly, what's your average?
I average about 15.5 for 3x3. Here are some thoughts comparing Corner Bias to Edge Bias.
Viewability:
When looking for the corner first you can see at most 12 of the 24 corner stickers (assuming you view the cube slightly at a diagonal). This is 1/2 of the possible stickers a corner be in place of, or 50% sticker viewability. Assume here that we have solved the cross, and have not yet solved any F2L slot.
For edges you can see at most 10 of the 16 possible stickers, viewing at a diagonal, where an unsolved edge could be. This is 62.5% sticker viewability.
Before I classify the edge bias as better than corner bias now let's look at how many of those stickers determine which piece you have.
Corners:
Of the 12 visible stickers, 1 corner has all 3 stickers viewable, and 3 corners have 2 stickers viewable (allowing you to determine which corner you have). This means that you can correctly identify 4 of the 8 corners at a glance without a cube rotation.
Edges:
Of the 10 visible stickers, 3 edges have both stickers visible, and 4 edges have 1 sticker visible. Notice that I am ignoring the cross edges, if that was not clear already. This means that at any given time you can correctly identify 3 of the 8 unsolved edges.
From a "piece" standpoint I would say that Corner Bias is better because you can identify 4/8 pieces or 50% of the pieces at any given moment. For Edge Bias you can identify 3/8 unsolved edges or 37.5% identifiable at any given moment. Of course I mean here identifiable without a cube rotation.
----
Now consider when you have solved the first corner edge pair at FR. If you view the FR slot as being right in front of you, with the cube on a diagonal, then you now can identify 3 corners of the 7 unsolved, and 2 edges of the 7 unsolved. This places the corner visibility higher than the edges for identifying pieces. For sticker visibility for corners there are 10/21 visibility or 10 stickers visible of the total 21 unsolved stickers. This is 47.6% sticker visibility. For edges there are 8/14 visibility or 8 of the 14 unsolved stickers are visible or 57.1% sticker visibility.
Sticker visibility seems to prefer Edge Bias, but visibility taking into account ability to identify pieces without a cube rotation prefers Corner Bias.
I would say, in my opinion, that I would prefer to be able to identify pieces without rotating, thus Corner Bias, than to see more of the remaining unsolved stickers when using Edge Bias. So in short, I think the visibility argument shows that "Corner Bias" lookahead is better if you prefer ability to identify pieces without a cube rotation. Maybe some experts use Edge Bias, and can argue an advantage over seeing more stickers? The downside here is that you must rotate more to identify pieces.
I will try to look into the number of stickers visible by not just looking at a side of the cube in the static sense, but looking at the cube through the flow of a 90 degree cube rotation of either y or y'. This would be something that all solvers do, and is very important to lookahead as well.
This post is already very long, so I'll stop here though and post the rest later.
Chris
