Train
Calibrate the lab
These answers set your starting depth only. From there the engine moves on what you actually do, not on what you claim.
Used only to personalize the coaching prompts you can copy out. Stored in this browser, never uploaded.
Every question here is about reasoning rather than about a specific patch, so all of them apply across the series. This only tailors the resource list and the coaching prompts.
Progress
Mastery is tracked per skill, and it is capped by the hardest tier you have actually cleared. You cannot reach Mastered by getting easy questions right repeatedly.
Your saved answers
Everything you type stays in this browser on this device. Export it if you want to keep it, study it, or hand it to a coach.
Review Protocols
A protocol is a repeatable procedure you run on real footage. Review without a procedure collapses into re-watching yourself play, which feels like work and teaches almost nothing. These unlock by depth, because each one assumes the ones below it are already habits.
Practice Plan
Generated from your two weakest tracks and from what motor-learning research says about how practice becomes performance. Regenerate whenever your progress shifts.
AI Coach
This page has no model behind it — it is one HTML file with no server. What it can do is build you a calibrated tutor: a prompt that carries your current depth, your weakest tracks, your archetype and your character into an AI assistant, and instructs it to teach the way this module teaches. Socratic, no answers handed over, one question at a time.
Resources
Filter by what you are actually working on. Community sites, creators and tools come and go — verify anything that looks stale before you build a habit on it.
How It Works
The four tiers are a thinking ladder, not a rank ladder
Every question is tagged to a tier. The tier is not “how good you are at Smash” — it is what kind of cognitive work the question demands of you. A locals-winning player who has never interrogated a habit can be stuck at Tier 1 thinking, and a mid-ladder online player who reasons carefully can operate at Tier 2.
| Tier | Demand | The question sounds like |
|---|---|---|
| 1 — Identify | Name what happened, and separate what you observed from what you concluded. | “What actually took that stock, as opposed to what I felt took it?” |
| 2 — Explain | Trace the cause through the systems of the game: space, percent, option coverage, frames. | “What made that stock unwinnable ten seconds before I died?” |
| 3 — Evaluate | Judge a decision on the information available at the time, not on how it turned out. | “Given what I knew then, was there a better option — and could I have executed it?” |
| 4 — Model | Reason about your own reasoning, and about an opponent reasoning about you. | “What is my model of this player, how would I know if it is wrong, and what do they think I am about to do?” |
How the engine grows with you
- Each of the eight tracks carries its own depth value from 1.0 to 4.0. Answering well pushes that track up; missing pulls it down. Depth falls faster than it climbs, which keeps you near the edge of what you can actually do rather than comfortably inside it.
- The engine serves your weakest unmastered track most often, so your time goes where the return is highest — with a deliberate minority of questions from stronger tracks so they do not decay.
- Some questions are archetype-specific. Where the failure mode genuinely differs by character type — approaching versus being approached, for instance — the engine prefers the version that matches what you main. Most of the material is deliberately archetype-neutral, because the reasoning transfers between characters even when the specifics do not.
- Mastery is capped by evidence. Proficient requires clearing a Tier 3 item. Mastered requires Tier 4. Grinding Tier 1 plateaus you at Familiar, which is the honest result.
- Items you get wrong are returned to you later, not immediately — spaced by several questions, because re-answering something you just saw tests your short-term memory rather than your understanding.
- Hints are a ladder: the first reframes the question, the second narrows the field, the third nearly answers it. Using hints reduces credit, because a hinted answer is weaker evidence that you can do it unaided.
Why the content avoids specific frame data
You will notice this module almost never quotes a frame number, a kill percent or a character matchup. That is deliberate. There are dozens of characters and thousands of interactions between them, so any table printed here would be both incomplete and beside the point — and looking your own numbers up is the skill. A player who knows their opponent's fastest out-of-shield option because they measured it in training mode is doing analysis. A player who memorized a tier list is doing trivia. The reasoning is what transfers between characters, between matchups, and between games in the series; the numbers are homework you do against your own eight or ten worst matchups.
What this module cannot do
Nothing here grades your written answers — there is no model behind this page. Free-response items are scored by an explicit rubric that you apply to your own answer after reading an expert response. That is a real and well-evidenced technique (self-explanation against a worked example), and it works, but it runs on your honesty. Every item also has a Copy coaching prompt button that packages your answer, the scenario and your current depth into a prompt for an AI assistant that can read and respond to it. The AI Coach tab builds a fuller version of that.