What an n-gram actually is
An n-gram is just a short run of n characters. A two-letter run is a bigram (th, er, in); a three-letter run is a trigram (ing, the, ion); a four-letter run is a tetragram (tion, ther). The "n" just stands in for the length.
Every word is a chain of overlapping n-grams. faster is the bigrams fa · as · st · te · er, or the trigrams fas · ast · ste · ter. Break any text down this way and you stop seeing 26 letters and start seeing a much smaller vocabulary of sequences you type over and over.
Why your speed lives in the sequences, not the letters
Once you can hit every key, individual letters stop being the bottleneck. What slows you down is the handoff from one key to the next, the transition. Some transitions are smooth because they alternate hands or roll across neighbouring fingers (th, er); others force one finger to double back or the same hand to contort (ol, my, br).
Fast typists aren't moving each finger faster. They've turned the common sequences into single motor gestures, the way a pianist plays a chord instead of four separate notes. That's chunking: your brain stops sending "i, then n, then g" and starts sending one "ing." N-grams are the unit that chunk is built from.
See which transitions are costing you time on your progress page →
Why drilling the common ones pays off
Language is lopsided. A small set of n-grams makes up a huge share of everything you type, th, he, in, er, an, and a couple dozen others appear constantly, while most possible letter pairs almost never do. So practice does not divide evenly across the keyboard: nailing the top few dozen sequences pays off on nearly every word, while a rare pair you fumble barely costs you.
That's the whole efficiency argument for n-gram drilling. Instead of retyping full paragraphs, most of which is sequences you already own, you do short, dense reps on exactly the handful of transitions that are slow. A minute spent smoothing ol and br buys more real speed than ten minutes of general typing that only touches them a few times.
It works because of how motor memory is built: repetition close together. Drilling packs dozens of reps of the same move into a short window, which is exactly the condition under which a sequence turns into a chunk.
How TypeCafe drills n-grams
Grams mode is built on this idea. You pick a source (bigrams, trigrams, tetragrams, or whole words) and a scope (the top 50, 100, or 200 most common sequences). TypeCafe then feeds you those sequences in short levels rather than random text.
It's a progression, not a static drill. Each level has a speed and accuracy target; clear it and you advance to the next sequence, with your running WPM tracked across the ladder. You're never grinding, you're climbing through the exact sequences that carry your typing, one mastered chunk at a time.
Then drill your own weak sequences
The top-50 list is a great start, but your slow transitions aren't everyone's. After a test, TypeCafe reads your keystroke timeline and finds the transitions you lose time on, then builds a drill saturated with exactly those pairs. That's the difference between practising common sequences and practising your costly ones.
Then re-measure. Run a normal test again and watch the transition drop off your weak list and the WPM delta move. If it moved, the drill worked; if not, drill something else. That diagnose → drill → re-measure loop is the whole point, n-grams are just the material it runs on.