Let's get to work. The percentages I have function as some way to quantify relative goodness in a given scenario. Of course, each scenario is specific as a case and should be treated that way. With that in mind:
This approach seems to ignore that mixing up options can be more effective overall than using the most effective option on its own. Let's say you approach with move A 50% of the time, and move B 50% of the time. A works 90% of the time, and B works 80% of the time. Why would you ever use B? Using B forces your opponent to defend against both A and B. If A is really good against shielding and the only thing the opponent can do to stop B is shield, then using them in combination could very well yield a higher success rate than using A or B on their own. If you only use B, they will just shield all the time, and if you never use B, they will never shield (making A less effective).
You do use B, it simply varies with the situation. The important part here is that you understand why B is better before you use it (there must be some way that B becomes your "best option"). I'm aware of how circular this can be, just to be clear on that point. It is your judgment at a moment's notice that determines which option is better and why. This is a skill that can be practiced and eventually mastered, and will be a great source of margin for you to use against your opponents.
In short, I'm not saying "only use A, never use B". I suppose the best way to put it would be "use A when A is better, and use B when B is better, and seek to understand why which option is better at any given time". Yes, this is obvious on paper. No, people still don't do it in practice. Perhaps articulating it will fix that.
I'm with you up until you start giving percents. What contributes to the percentage effectiveness of a given tool? Is it how consistently it works vs that character's hitboxes? How well a player can use it and abuse the threat of it? Is it because of the move's statistics(startup, cooldown, IASA frames, etc)? I agree we should start with ruling out what GENERALLY doesn't work and what GENERALLY does work and of course WHY we believe those things to be true, but to say a tool is statistically better over another gives me great cause for hesitation, ESPECIALLY given that opponents can expect you to use one tool over another if you begin thinking this way.
This whole mess is my way of defining "don't do bad things". What makes something bad? To me, it is anything that unnecessarily lowers your win rate. We rule out what generally doesn't work because it is readily obvious. What is not obvious is to also rule out things that don't work work as well as another option, because both of them are still bad but to differing degrees. The latter is just not as obvious, but it's still going to affect your performance and it would be in your best interest to fix it.
I know this is a magnet for trolls to say "Umbreon is giving obvious day 1 advice" but this is something that really happens at every level of play. You still pick sub-optimal options too, although they are either much less frequent than other players, or my lack of knowledge with Falco allows me to miss them. But when you switched to Marth in GF, oh yeah I can accurately judge every single action you did and that's why Sensei knows that you still don't get a 100% yet even if you beat the best player in the world. For lower level players, sub-optimal decision-making is much more common and openly apparent even without a mastery of character knowledge.
One tool is almost always going to be objectively better than another. The player may not understand why, which tool is better, or how to reason through that type of judgment in a very small amount of time. That's okay. We can talk about it on discussion boards and test and practice to resolve those issues. What is not okay is to know that you have a better option and to intentionally not use it. It happens
all the time.
Efficiency, to me, is knowing your BEST options(let's say anything over 75% success rate since you like percents) and knowing how to harmonize them with movement(if we consider that separate from tools which I think we should since it's what tools come from and should be worked on its own) to produce the highest possible rates of reaction time, difficulty to read, and hopefully ease of inputs....which all sounds like what I'd say my definition of efficiency is anyway. To say Dtilt is ALWAYS 80% or grab is ALWAYS 90% is silly, regardless of good movement setups but especially factoring in opponent awareness(I'm aware your example was thrown together but I believe my point is still valid even in-depth.)
Bingo. You are totally valid, although you should change your percentage to 100% and play to the ideal. You're just jumping too many steps mentally because you already know this stuff I think. For example, I think we should eliminate predictability from our evaluation for now, something you alluded to in your post. We can add it later once we understand our options better.
I find it hard to apply this mentality to moves when I feel there are few situations where you could directly compare two moves solely on success of being landed. You would agree there are other factors to consider V move over U move due to if it lands, you end up in a more favorable position or not.
There's many, many variables that defines one option as "the best" and knowing what your best option is should be MUCH harder than simply using it. One step at a time. Once everyone agrees that yes we should be sticking to our best options we can start to figure out what those options are and why they are the best.
edit: speaking of armada so much, here's what I was listening to while I wrote this post. Let's see who gets it.
http://www.youtube.com/watch?v=_rHYRaJGDk4