It has been explained again and again, that it isn't calculated like that. It is a comparison between the number of infections between an unvaccinated group and a vaccinated group.
If you have 1000 unvaccinated people and you find that 200 got infected, a 90% efficacy means that out of a similar group of 1000 vaccinated people you would expect 20 of them to get infected. Engaging in high risk behavior, or a more transmissible variant increases the incidents of infections. So 1000 unvaccinated people engaging in this scenario might lead to 500 infections. Then a 90% reduction would mean you would expect 50 infections in vaccinated people in a similar environment.
EDIT: And I'll add, that while I used infection in the example above, when we are talking about the COVID vaccines, we're usually talking efficacy versus severe outcomes. Versus avoiding infection entirely or symptomatic infection, we should expect it to be lower. How much lower we'll find out as these next waves blow through. The Israel study from today, is not promising. Singapore study had it 70% vs any infection, and 80-90% against symptomatic infection. The US should be revealing since it will be more of a baseline, since the US is essentially down to vaccines and not vaccines + other types of mitigation.