One thing is to say that a war is unlikely to be started by a woman (your historical observation); another thing is to say that a woman is unlikely to start a war (your claim). That's because, historically, the people in a position to start a war have mostly been men.
If W means that a given leader (i.e., someone who could have started a war if he or she wanted to) was a woman and S means that the leader started a war, by Bayes' theorem the probability of the leader having started a war, given that she was a woman, is
P(S|W) = P(W|S) P(S) / P(W)
Likewise for men (if ~ means negation):
P(S|~W) = P(~W|S) P(S) / P(~W)
You observed that P(W|S) is much smaller than P(~W|S); but to compare P(S|W) and P(S|~W) you also need to take into account how much smaller P(W) is than P(~W).