What the review added to a CS2 prediction after the match contains, and how to read it against the original call.
When a match is played, its page gains a review. The original probability, the picked side and the reasoning stay exactly as they were published. The page ends up carrying both halves — what was expected, and what happened — which is the only arrangement in which either is worth reading.
Right or wrong, marked plainly, at the top. There is no partial credit and no re-framing of a miss as a near-miss. The same marking is what feeds the published record.
A short account of how the match went and which factors mattered. The value of this part is that it can be compared against the risks listed before the match: when the thing that decided it was already on that list, the call was wrong for a foreseen reason, which is a different failure from being blindsided.
Picking the right side for the wrong reason happens, and the review says so when it does. Counting it as a win is honest — the record counts outcomes — but treating it as a validated method is not, which is why the account matters as well as the mark.
A match that finished has an unreviewed period before its review is written. During it the page shows the original call and no outcome. A missing review is a pending one, not a hidden one.
One review explains one match. A run of them shows whether the same kind of mistake keeps recurring — a map pool consistently misread, an online-versus-stage adjustment that is too small. The archive is where they can be read together.
Without reviews, an accuracy figure is a number a site chose to display. With them, every entry in the total has a page saying which match it was and how it went, and the figure can be argued with. How accuracy is measured sets out what goes into it.
It does not revise the probability, re-pick the match, or explain away the result. It records the outcome and what caused it. A review that argued the original call was really right would make the whole exercise pointless.