If you are missing points on SAT Reading and Writing, the comma rules are probably not where they are going.

We rate practice questions on a five-point scale for how hard they are, and the test we use is whether a strong student working quickly would be genuinely pulled toward a wrong answer. To check that our scale means anything, we mix in official College Board questions with their published difficulty labels hidden, and rate them the same way in the same batch. Reviewers see our questions and the official ones together and do not know which is which. That gave us something we were not looking for.

Punctuation questions run out of room to be hard

Take the questions College Board itself labels hard, and sort them two ways: questions about the mechanics of a sentence, and questions about what a passage means.

That grouping is ours. College Board sorts the section into four domains, and one of them, the revision questions covering transitions and rhetorical synthesis, belongs to neither of our two piles. We left those out of the comparison rather than force them into a group they do not fit.

The mechanics questions came out easier. Not by a little, and not only among the hard ones: the same gap appeared at all three of the labels College Board uses, easy, medium and hard.

This is a few dozen questions per level, which is small, and we are reporting a direction rather than a measurement. What makes it worth saying is that the direction did not wobble. Three levels, same gap, every time.

A hard comma question has one place to hide

There is a reason for this, and it is structural rather than accidental. A punctuation question has a small number of things it can do to you. The sentence can bury the real subject behind another noun. It can hang a modifier where two nouns could claim it. It can put a boundary between two clauses where you have to notice both halves could stand alone. Once you know those moves, you know them, and the question has nowhere else to go.

An evidence question can be hard in a way that does not run out. The passage can support a claim partially. Two choices can both be true of the text while only one answers the question that was asked. The right answer can restate the passage in words that never appear in it. There is no finite list to learn.

That difference shows up in what happens when a strong student meets each type. On the mechanics questions, reviewers frequently could not name any wrong answer a well-prepared student would seriously consider. On evidence and inference questions, they usually could.

We tried to write harder grammar questions and hit the same wall

This started as a complaint about our own work. We had been treating our grammar questions as underpowered. They sat well below where we wanted them, and we spent real effort rewriting the instructions we author them from, twice, trying to push them up.

Then we rated the official hard grammar questions on the same scale and they landed in the same place. We were not writing weak questions. We were writing questions that matched the ceiling the skill has. That reframed the problem from a writing failure into something about the material, and it is the reason this post exists at all.

What the number is not

This measures how much temptation a question carries, not how many students get it right. College Board's own easy, medium and hard labels come from student performance data, which is a different and in most ways better thing to know. Ours is a judgment about the answer choices. A question can be genuinely difficult because it is long, or because it appears when you are tired at the end of a module, and none of that shows up in what we measured.

So read this as a claim about where the traps are, not a prediction of your score on any particular question.

Spend the time on evidence and inference

The practical version is short. Learn the mechanics rules properly, because they are learnable and they are about a quarter of the section. Then stop. There is a point, and most students reach it faster than they think, where another hour on semicolons returns almost nothing, because the questions cannot get much harder than the ones you can already do.

The questions that separate students are the ones asking what a passage supports, what follows from it, and what a specific sentence is doing there. Those are slower to improve at. They are also where the remaining points are.

If you want to check this on yourself rather than take our word for it, go back through a practice section you have already done and sort your wrong answers into mechanics and meaning. Most students find the pile is lopsided, and it is usually not the pile they have been studying for.