There is a difference between assuming that they are expecting you, and having your enemy know your every.single.move.as.you.make.it.
EX:
If I'm going to punch you in the back of the head, I'll sneak up behind you. You might know I'm there, you might not.
Now, if I come up behind you at the same speed while yelling, "I'M COMING UP BEHIND YOU AND I'M GOING TO PUNCH YOU IN THE BACK OF THE HEAD," you'll definitely know I'm coming, and I should expect a countermeasure.
I think there's two conflated questions here. I should ideally plan my attack as if my opponent knows my moves, where possible. I should not tell my opponent my moves, where possible. Just because it's not on the news networks doesn't mean the enemy has no way of observing you.
If we're boxing, or another adversarial match, though wouldn't you become suspicious if I was seemingly unaware that you were behind me?
This isn't a "suprise! we're here to check up on things" this is something the kidnappers should be expecting and planning for (what else do they think will happen).
Boxing is kind of a great analogy here, but not in the way you think.
Boxers have this concept called "telegraphing" to call out patterns of movement that indicate what they're about to do. It's one of the easiest ways to lose a fight because an opponent who can read your tells can react to them instead of to the move that they precede.
Even though both participants know that the other is actively attempting to attack them, there's still a significant amount of strategic value in concealing the specifics of imminent activity.
The difference was that they knew exactly where the team would break in. The difference between "there will probably be people trying to enter the room" and "I'll shoot that guy once he breaks the window over there"
Expecting, yes. Watching your every move in real-time, in 1972, no.
Remember, this was in the early days of television - not the continuous video/metadata surveillance state we have now, with commensurate capabilities available to any idiot with a motive and few hundred bucks. Sure, the police assumed the terrorists were expecting a raid somewhere between right now & hours/days from now and involving a lot of heavily-armored police with lots of firepower, but wouldn't know which second and the exact arrangement & capabilities of forces involved. Even seeing all the TV cameras around, the police hadn't learned to make the instant & automatic connection between a video crew and the TV set on in the hotel room about to get raided. This was the first time the targets of a raid were able to see that there were, in fact, right now, 12 cops standing just feet from the windows, and 5 approaching the front door carrying machine-guns and about to bash in the door with a battering-ram in 7...6...5...4...
You should expect it and plan for it, but when you know it your actions have to change. If I know you're about to kick down my door, I have far more options than if I know "in the next hour they'll probably kick down my door". The information completely alters the dynamic. Even in games with no hidden information (such as chess or go), you don't know what my next action will be (except in few cases where my options are severely limited), you only know the probabilities. What if we're raiding a house for some reason (like in Munich), and there are 4 doors and a dozen windows. 5 gunmen inside, if they know which doors we're about to go through and which windows we're ignoring, they can either effect an escape or a more effective defense. Our raid is going to be fair less effective if we announce our plan to the opponent instead of making them guess at it.
Ah, but that's about cryptosystems. And its application is really about the continued use of a thing. Kerckhoff's principle doesn't apply so much to a one-off situation. Sticking with a house raid:
We have a plan to raid this house using these entrances, covering those exits with these 10 people. The adversary knowing our plan screws us, but learning it after the fact doesn't help our adversary.
But what if we use the same tactic every time? What if we always cut power immediately before breaching. That's a signal to the adversary. What if we always use the same sniper and he has a known handicap? If the adversary learns these things then they can break our efforts (or reduce the effectiveness) to conduct the raid. And this is where Kerckhoff's principle could be applied.
EDIT: It's not secrets that are the problem, it's keeping the wrong secrets. You can't have my passphrase or my PGP private key(s), but, sure, you can have the AES and RSA algorithms and my public key(s).
Sometimes a fragile plan is still the best one you can have. Life doesn't always give you good choices. Sometimes you're stuck with either a fragile plan where things can go horribly wrong if secrecy is breached, and one that's even worse.
