Comparison of OLL and CLS

I'm trying to determine if using CLS during step 4b during bad cases, actually yields a lower move count if I turn it into a CLS case, instead of fixing it and doing oll. Just for ease of comparison I will use the following case:

R U2 R' U R U' R' U

The inverse would be my solution to solve this case normally, but turning it into a CLS case makes the solution merely

R U R'

This is 5 saved moves, but I'm wondering if it really is 5 saved moves.

I can calculate the average move counts for both oll and CLS, but I cannot factor in which cases occur most often, and give an actual average move count for oll or cls. Can anyone help me?

You mean OCLL. What I don't get is, why would you want to learn 100+ CLS algorithms when you already know all 20 F2L cases (with the edge oriented) and all 7 OCLL algorithms?

I can tell you right now that the average move count for inserting the last F2L edge is 2.6 (with probability of occurrence factored in). You'd have to ask Lucas for the average move count for CLS.

Wow I really meant to say the subset of CLS I + Im which are the cases where the corner are in place but twisted. I'm sorry...=\

I use it for Petrus. "EJF2L"

So I decided I'm no longer using cls because the algs aren't as fast as my olls, and on average I found it saved about 1 move.

17, 22 = -5

17, 20 = -3

23, 20 = +3

17, 17 = 0

17, 15 = +2

16, 16 = 0

15, 18 = -3

14, 18 = -4

20, 21 = -1

15, 16 = -1

These were random step 4b cases I got to using RU scrambles, and solving Step 4a. I only used cases I would normally turn into cls cases, and I solved the case with CLS, counted it (left), and then undid it, and solved it as I would with normal ocll. I took the average difference of the move counts and found it was 12/10 moves saved. But then again I guess it's not exactly hurting my times.......eh. I'm not sure what to do. It can be really helpful if I get a really really nice case for it, like a 2x2 with a twisted corner. Okay, I'll keep it but I'm no longer advising those I teach to learn it as in my opinion it is not worth learning.

When you constrain your cases to the I group, their relative probabilities stay the same as they were in full CLS. I0 occurs four times as often as I7. I0=4/27. I1=4/27, I2=2/27,...

The average move count of I is the sum of (# of moves)*(prob. of occurence) for each case.
(4*(11+9+11+9+9+11)+2*(11)+1*(13))/27=275/27=10.19

The move count for full CLS is listed as ~9.5.

I believe that it would be quicker just to do OLL after as it is easy as to recognize (since you are doing 2-gen stuff anyway) and you should be able to get all of the sub 1.5 and a few sub 1.