It seems to me that the #1 way to predict pilot performance in emergencies is whether the pilot has encountered the emergency before and lived. If so, running

It seems to me that the #1 way to predict pilot performance in emergencies is whether the pilot has encountered the emergency before and lived. If so, running the pilot through the most scenarios possible in a given unit time would seem to be a good strategy to train pilots, and simulators would increase the rate of scenarios per unit time versus a real aircraft. However, you can't get away from actually flying a plane. There's little things you don't practice in the simulator, like making sure you have your handheld VHF radio, checking for water in the fuel tanks, and how to balance cargo, that you'll never learn in front of a computer screen. If it's anything like network administration, I'd say something like 20% real experience vs 80% simulated might be a good mix. People who favor just one or the other lack performance under adverse circumstances compared to admins who do both, I've noticed.