TLDR: You are the only HN genius who forgot to multiply by allowed collateral damage.
They suggest based on thin evidence that it selects 90% Hamas associates given relatively clean data at start of conflict. There are oh so many things wrong with this.
- Hamas is the local government. Those who don't participate in the fighting aren't lawful targets in the first place
- They preferentially strike homes during the night maximizing collateral damage in order to obtain a higher chance of killing the target. They set the acceptable losses at 20–100 based on rank and importance of the target.
- Their initial accuracy was assessed by vetting a small sample earlier in the conflict.
What happens to your targeting as the conflict proceeds? Your known targets die, flee, and move around. New soldiers are recruited but don't provide clear intelligence from a chaotic warzone of their present status. You would logically expect such a system's accuracy to decline towards randomness as such a conflict proceeds and intelligence and targets become thinner on the ground. There is no reason to accept the initial 90% targeting accuracy on faith.
Even where we accept we must not forget to multiply by acceptable collateral damage.