I don't mean to confuse. I am sure there was a point when zwift relied on a power curve. This is because of the way bkool broadcasts power over ant( proprietary at the time like their bluetooth implementation is now). I have spent the last 20m running through Zwift posts about the pro but cannot find the post i saw about this. TR were having the same problem with bkool. zwift had a pro, they had to wait for bkool to release to them the Ant+ fec code/whatever for proper power readings and ERG mode (for training). I think it was only then that power/ERG broadcasts were properly readable in TR and Zwift. So, it probably is broadcasting power or a message where power is approximated and, interestingly, more accurately for gradient/transition in zwift than in Bsim.