Bet sizing carries information independent of the cards: a small bet keeps a merged range in the pot (value hands and marginal hands betting the same size, because the size is cheap enough that even a moderate hand's opponent-facing math still works), while a large bet polarizes the range (only the strongest value hands and the best bluffs can profitably fire it, because the size demands a lot of fold equity — see the MDF topic — or a hand strong enough not to need any).

The practical version of that idea is sizing to purpose. Before picking a number, name what the bet is for: deny equity (make live draws and overcards pay a wrong price to see the next card), get value (get called by worse), or build leverage (set up the remaining stack so a later street can be an all-in that makes sense). A size chosen out of habit — "half pot, because that's what I always bet" — serves none of those on purpose, and gives away information for nothing whenever the habit doesn't fit the spot.

Two board contrasts make it concrete. On a dry K♦7♠2♣ after you open and the big blind calls, the texture favors your range — you hold the overpairs and the strong kings, the caller mostly doesn't, and there are no draws to charge (the Board texture topic's walk of this exact board shows the class-by-class detail). The purpose is cheap, range-wide pressure: his unpaired hands — most of his range — can't profitably continue against any size, so a small c-bet of a quarter-to-a-third pot buys nearly the same folds as a big one while risking a fraction as much, and it lets your whole range bet the same way without telegraphing which slice this hand came from.

Flip the texture to 9♥8♥7♥ and the purpose flips with it. Hands strong enough to bet now genuinely need protection and value at once: draws are everywhere, they're paying to outdraw you, and a big bet — two-thirds pot and up — is what makes their price wrong. But the medium hands that liked small-betting the king-high board don't size down here; they check instead, because betting builds a pot exactly when a scary card is most likely to arrive. That's why big-bet ranges polarize (the strongest hands plus the best semi-bluffs, nothing in between) — the size itself filters out the middle.

This app's postflop grading model implements the same merged/polarized split as a small set of range-narrowing constants — printed in full, live from the shipped parameter set, on the How this app grades you page.