Better CV will be (for me).
720 normal CV
684 CV (after 5% lost due Fatigue).
This should be on the Counter. The other values are random. And attacker die roll should not be count! This is important, because it changed the defensive values, too and it didn't happend, if you defend.
In your example:
experience and moral:
Propability that you double on one is 3/4 = 75% (right?), that nothing happens is 1/4 * 3/4 = 3/16 = 18,75% and 1/16=6,25% that your cv will be halved.
You now had results: your cv is 171 (probability is 1/256), you cv is half 342 (probability is 6/256), you cv is 684 (no change, probability 9/256+6/64), 1368 (double, 18/64) and your cv will be 2736 (4 times, probability 9/16).
This means 56,25% that you will have 2736 CV and 28% that you will have 1368 CV. This means 84,375% that you will have 2736 CV or 1368 CV.
You had 1268 CV. (by the way, you went wrong there with 0.75 multiplier instead of 0.95).
So my suggestion is, give the cv*fatigue multiplier as CV on the counter. Than you know what could happen.
Maybe I made a mistake with fatigue. But I still thing using expected value is better than completely ignoring random factors.
If you have base value of 10, 50% chance to double it, and 25% chance to halve it, then expected value is 0.5*20+0.25*10+0.25*5=10+2.5+1.25=13.75. Random value calculated in a lot of rolls (and CV of all units involved in a fight depends on hundreds of rolls) usually floats somewhere around that expected value, though in rare cases it will of course differ a lot. Expected value is much more accurate than just pure base value. A unit with the same base value but just 25% chance to double it, and 56.25% chance to halve it will have expected value of 0.25*20+0.1875*10+56.25*5=5+1.875+2.8125=9.6875. So one unit would have a value of 14, and the other of 10, instead of both units having a value of 10. This is much more accurate assessment, and the whole purpose of "better CV" mode.