I think making uncalculated sacrifices is a very good way to lose (believe me, I do it all the time). You are intending to limit prediction with HO, but making obvious ones continues to be a useful play (assuming your current Poke has enough HP left to be worth saving and/or is fast enough that getting one additional attack out of it later is probable- there's no need to retain guys that are actually useless unless the switch-in doesn't take any damage, i.e. is immune). The point of sacrificing is to:
1) Kill off the one Poke that walls your team, opening up a sweep (e.g. if Rotom is the guy that stops your team - as he so often is - sacrificing a monster or two to finish him off is acceptable if it results in you being able to set up Scizor or Lucario or Gyarados and tear the rest of their team apart).
2) Avoid mispredicting or taking unnecessary damage in situations where you can follow up with a sweep. This is most often to scout for Choice items, both to see if they have one and to see what they lock into. It's important to note that it is not always necessary to sacrifice this way to guarantee that they do what you expect. A lot depends on how reasonable it is for them to make the non-obvious move, and the severity of the consequences for you if they do, and the significance of the reward.
3) Least exciting, you often sacrifice your support Pokemon for the purpose of shielding your attackers for as long as possible. Nothing fancy here.
At the risk of giving really obvious advice (or worse, being wrong in my analysis), I'm going to toss out an example here. Say Scizor comes in vs. your weakened Starmie. He's almost certainly going to Bullet Punch, so you could go to Empoleon right away to try to save your frail monster. That is almost certainly wrong, because on the off-chance that they U-Turn or Superpower, your Empoleon plan is going to fall apart. You are better off sacrificing Starmie to make sure they're locked into Bullet Punch first, then going for the sweep when you know they can't stop you. The sacrifice in this situation is the safe and prudent play, not a reckless one.
In this example, Bullet Punch might be the most likely, but U-Turning is still a sensible move for them, and if they U-Turn, probably you will never get Empoleon off because it relies on catching people off-guard. Even Superpower could be reasonable for them, as long as they kill Starmie anyway, since it protects against switched-in Magnezones, and obviously this causes disaster for the penguin. Meanwhile, the cost of sacrificing a weakened Starmie is minimal (particularly since Empoleon after an Agility is essentially a better Starmie anyway), and the reward for setting Empoleon up (by the time you're considering it) should be a prompt victory.
On the other hand, if you had a full HP Starmie out and they brought in Scizor, they're much more likely to U-Turn to avoid locking into Bullet Punch, because they're going to have to take a hit from you either way (Bullet Punch doesn't kill you, and you don't kill them either). In this situation, there's little to be gained from sacrificing Starmie (unless you absolutely need to beat up Scizor for some reason), so you don't need to sacrifice, you just play normally and send in a U-Turn resist.