I want to say why the cardinal systems (STAR/Score/Approval) are usually better than the ordinal (Ranked/RP/Schulze). It is all about ,,embedding''. More efficient embedding systems mean you can get a higher information gain. Let's look at two different elections but with the same premise.
Suppose you REALLY like both candidates A and B. Then suppose that you really dislike candidate C.
Ranked:
A > B > C (A/B position determined by flipping a coin)
Ordinal:
A:5 B:5 C:0
What's interesting here is that the first (ranked) system tells us (embeds) that our preference for A over B is the same as our preference of B over C. But in the second case we are able to tell the system that we actually like both A and B equally and that the distance to C is much larger. Ranked does not capture how we don't like C and doesn't capture that we do like B. Because of this the ranked systems frequently fail the favorite betrayer criteria (RCV fails btw). In a realistic scenario this means that in a ranked system Bernie would spoil the election for Biden (or vise verse). Although Gary Johnson wouldn't be a spoiler in either Ordinal nor Cardinal I'd argue that a weak spoiler is less of a worry than a strong spoiler.
I will also argue that Ordinal methods are far simpler to calculate and do statistical verification on. Ranked systems require a while loop where we pop an element from our array each iteration and reallocate that element's votes until there is a singular winner. Cardinal systems on the other hand are simply an sum argmax. This sum argmax can be done in parallel and also makes it far easier to verify elections with random subsets of votes (a fairly common practice btw). The simplicity of the algorithm used to tally votes is extremely important and increases the transparency of the election.
At the end of the day, the two systems are
Ranked: "Put the candidates in order of preference"
Cardinal: "Rate each candidate from 0 to 5 stars"
People are very used to both types of questions. I see several arguments saying the latter is more complicated but this clearly isn't true or we'd have to change the vast majority of surveys.
I also want to state that there are no globally optimal voting systems. Yes, there are plenty of flaws in cardinal systems. But because no global optima exists we have to instead argue that the tradeoffs are worth it. There are some objectively better systems than others though. To play fair like this I will say that I believe that the above advantages are worth the slight reduction of VSE in a 100% honest election that you'd obtain from using either RP or Schultz style elections.