Week 1: diagnose and build your buckets
Take one full Reading and Writing section, or use the most recent one you took, and do the step almost everyone skips: sort every miss into a bucket by question type. Punctuation. Transitions. Words in context. Command of evidence. Inference. Central ideas. Just tally them.
You expect fifteen scattered problems. You will almost always find two or three fat buckets wearing a costume of chaos. That is the best news of the six weeks: "raise my score" is vague and paralyzing, while "fix my comma rules and stop switching answers" is a to-do list. (The full sorting method lives in the plateau playbook.)
Then split each bucket in two. For every miss ask one question: would I get this right if I saw it fresh tomorrow? No means a knowledge gap, so learn the thing. Yes means a behavioral leak, so you knew it and lost it anyway. Those two need completely different fixes, and mixing them up is why studying hard moves nothing.
Self-test to run today: pull your last section, give yourself ten minutes, and tally the misses into buckets and into knowledge-versus-behavior. If you have never done this, it is the highest-value ten minutes on this page.
Do Week 1 in ten minutes, free: start the diagnostic at simplesat.study sorts your misses into buckets and gives you an honest read on your behavioral leaks. No signup, works as a guest.
Weeks 2-3: drill the buckets, grammar first
Now drill, one bucket per short session, and start with grammar. Standard English Conventions is about a quarter of the Reading and Writing section, and it is the fastest bucket to move because it is rule-based: there is a right answer with a nameable reason, not a judgment call. You are not memorizing forty rules, you are learning a handful of mechanical checks. The Period Swap settles semicolons and comma splices: swap the mark for a period, and if both sides survive as full sentences the boundary is right. The Delete Test settles commas around extra information: delete the middle, and if the sentence still stands, the commas belong. The Cross-Out settles subject-verb agreement: cover the phrase between the subject and the verb so a nearby plural noun stops fooling your ear.
Original example for the Cross-Out: "The box of rusted tools and tangled wires ___ in the shed." Cross out "of rusted tools and tangled wires" and you are left with "The box is in the shed." One box, one singular verb. The interrupting phrase was the whole trap.
Work one rule family per day, fifteen to twenty questions, then mixed sets. Two rules from learning science to obey. Spacing beats cramming: twenty focused minutes a day beats a four-hour Sunday at equal total time, because the gaps between sessions are where the skill consolidates. And re-solving beats re-reading: after every miss, close the explanation and rebuild the answer from scratch out loud, including why each wrong choice is wrong. That is the Rebuild; "oh yeah, that makes sense" is only recognition, and recognition evaporates. Cap each miss with a One-Sentence Autopsy naming the real reason; "careless" is banned.
Once grammar feels automatic, move to your reading-method buckets. Transitions are the fastest of those (one method, small taxonomy), followed by words in context. Spend the back half of these two weeks on whichever reading bucket is fattest.
Week 4: the behavior layer
Some misses are not knowledge at all. They are leaks: points your process loses even when you know the material. No amount of grammar drilling touches a habit, so Week 4 hunts them. Three big ones, each with a self-test and a fix.
First-instinct override: grade your last two sections twice, once on your first picks and once on your finals. If your firsts win, your reviewing is talking you out of right answers. The fix is the Receipt Rule: change an answer only when you can point to the exact word that proves your first pick wrong. A feeling is not a receipt.
Anchoring: on ten questions, cover the choices on the first five and commit to an answer in your own words before you look. Compare accuracy to the five you worked normally. If the covered half scores better, the choices were pulling you off your own read. The fix is Predict-Then-Peek: decide before you look, every time.
The time-sink: list every question you spent 90-plus seconds on, then look at the three questions right after each. Those downstream misses, the ones that feel "careless," are where the time-sink actually costs you. The fix is the 90-Second Flag: if a question has not clicked by ninety seconds, flag it, guess, and move. (All three in one place: the behavioral leaks guide.)
Week 5: two timed simulations
Now, and not before, bring back the full timed test. Two of them this week, spaced (one midweek, one on the weekend), under real conditions: two 32-minute modules, 27 questions each, Module 2 adapting in difficulty to your Module 1 performance, no penalty for guessing so never leave a blank. These tests integrate what you fixed and build stamina; they do not re-measure what you had not fixed yet.
Then review each one properly, which takes about as long as taking it. Rebuild every miss, autopsy it, and drop it back into knowledge or behavior. (How to review a practice test is its own protocol.) Judge the two scores as a pair, not as a verdict on a single day: College Board reports section scores as a range rather than a single point, because scores vary between administrations, so a small gap between them is noise rather than a trend.
Week 6: go light, and sleep
Taper. The week before the test is not when cramming adds points; it is when sleep and a calm nervous system protect the points you already have. Keep your hand warm with one short mixed set a day, about fifteen minutes, no full tests. Reread your own autopsy notes so your named fixes stay loaded: run two-pass triage, honor the review screen, keep the Receipt Rule ready. Confirm logistics early so nothing on test morning is a surprise, then sleep, especially the two nights before, not just the last one.
The order is the part that matters most
Most stuck students run this plan backwards: simulate, simulate, simulate, then wonder why a month of Saturdays bought almost nothing. A full test measures you; it does not teach you. Diagnose tells you what to fix, drilling fixes it, the behavior layer catches what drilling cannot, and simulation proves it under a clock. Diagnose, drill, behavior, simulate. Run it in that order and six weeks is enough; run it out of order and six weeks becomes six measurements of the same unfixed leaks. With less time the sequence compresses (see the two-week plan), giving each phase fewer days, never a different order.