I have done this test many a time in the past and people may argue that the ERG mode will eliminate any “mechanical” changes one makes to the gearing. However, and this is my findings (I am going to use fictitious numbers just to provide some guidance). Now again, this is my findings and I am by no means an expert: I used to ride single speed and I then started using the ss on the trainer, the gear ratio was a 32T in the front with a 14T in the rear. I then started running out of resistance so I changed it to an 11T in the rear, same thing happened so I went with a 38T in the front.
Fast forward a couple of years where I’ve gone from the ss over to some gears (1x9) with the 38T in the front.
The resistance with this config works fine except on the high end of the spectrum. If I do a video with lots of sprints or let’s say the 5 sec sprint of the 4DP, I spin out, my trainer has a max resistance of 900W and I can only push around 600 before hitting a 150 cadence. So I then decided to go and increase the 38T to a 52T in the front and 10 in the back.
The result: My cadence is always lower than what the video requires, even if I go lighter in the back and I can push more watts on the sprints. So yes, ERG mode does provide resistance and it is suppose to make no difference when you fiddle with the mechanics of your gears but, the ERG mode has a band in which it operates (technical gurus please correct me if I am wrong) but my understanding is that if you increase the resistance of the gears enough then you can essentially override the ERG function and get more suffering for your effort