Let's assume, for the sake of argument, that someone would want to get Help on an attack only when they had a low chance of hitting, similar to our scenario. Let's further assume that "low chance of hitting" is 20% or needing a 17+.
In our scenario of 6 attacks needing a 17+ to hit means that, in a random trial of 100,000 attacks you'll hit this often:
0: 26,132
1: 39,250
2: 24,776
3: 8,181
4: 1,497
5: 156
6: 8
Average number of hits: 1.20
If you attack 3 times at advantage then you get the following distribution (using a different set of 100,000 attacks. It should probably be the same set but I don't want to redo things again and the difference is minimal)
0: 26,184
1: 44,158
2: 24,936
3: 4,722
4: 0
5: 0
6: 0
Average number of hits: 1.08
There are only six attacks so it probably wouldn't be too hard to mathematically figure out exactly what your chances of getting X number of hits would be. It's just that I already had a Monte Carlo simulation up in Excel so it was easier doing it the way I did.
EDIT: One commenter who mentioned it wasn't beneficial using the Help action made sure to add the caveat that the attacks needed to be identical in hit chance and damage in order to definitively state it wasn't useful to use Help. As you can see, in a low hit chance environment the Helped attacker only needs to do a little more than 10% more damage for it to be worth Helping.
For example, if you average 1.20 hits that do 1 point of damage each you'll do an average of 1.20 points of damage a round. The Helped attacker needs to do 1.20/1.08 or 1.11 points or more damage per attack to make Helping an equal or better option.