Im doing a security course, and a lot of swat police and whatnot use people in my course for general testing.
we we're given rubber knives with felt on the edge, and told to attack the (trained) police and try to "kill" them. they had to disarm us. we did it about 100 times. and out of that, every single officer got cut, and 60% of them we're cut fatally.
Then we did something with replica glocks which shot paint, they would start 5 metres away, charge us and try to get the gun. (none of us have gun training) anyway, at close range like that, we hit the officer about 40% of the time. at over 8 metres we could not land a hit except by chance.
Point being, for what we did, at close range, knives are far more dangerous than guns, and I would question any knife disarm techniques safety and effectiveness, as for guns, if someone points one at you who doesnt know what they are doing, theres a fair chance of being able to leg it if your over 8 metres.
Interesting stuff to consider and weigh up
One thing that bothers me about this type of data gathering method is that the rubber knife attacker attacks with 100% because there is 0% chance of actually hurting their victim. The defender only defends with maybe 60% at best because they are only roleplaying and don't want to really hurt the rubber-knife attackers. I don't know how the statistics would change if the defenders actually fought for their lives (not holding back)using well rehersed and effective technique. I would say, however, I wouldn't want to be attacking them with a rubber knife if they were...