What happens when you run an office-shooting simulation, with two tactical trainers, armed with ARs, as the shooters, and random volunteers as the victims? Watch the (NSFW) video:
In a previous simulation of a school shooting, they found that an armed defender was almost always able to either kill the attacker or to prevent them from entering the classroom and killing more students.
In this case, not so much. Two trained attackers, operating as a team, are more than a match for one untrained defender with a handgun — most of the time:
In one of the early scenarios, a relatively new shooter decided that instead of trying to confront the armed terrorists she would use her gun to cover her retreat and give her co-workers time to escape. This plan worked perfectly, and she was able to escape from the room while returning fire towards the attackers, allowing nearly everyone in the room to escape before she too turned tail and ran.
In the face of overwhelming numbers and firepower, it appears that this tactic using the firearm as a means to give everyone else time to escape is extremely effective. There was only one person who used this tactic, but they used it to great effect.
Apparently there were serious issues with the masks provided to the victims/defenders fogging up in the cold, such that they were mostly able to shoot only in the general direction of the attackers. TTAG has a pretty bad reputation in the gun community (plagiarism, attention-hounding, etc.), and they ordered this scenario set up in only 24 hrs which forced the company performing the tests to cut a lot of corners.
I think the larger problem is that the trained attackers knew to methodically clear the small building against some (light) resistance.
Also they are assuming a suicide attack, whereas the actual attack wasn’t. It makes deterrence far more powerful if the attackers plan to survive.