UVM scoreboard
A UVM scoreboard is the referee that decides whether the design did the right thing. It receives the transactions a monitor observed coming out of the design, and compares them against what a 'golden' reference model says should have come out — flagging any mismatch as a bug. Picture a chess arbiter holding the rulebook: every move the players make is checked against what is legal, and the instant something illegal happens, the whistle blows. The scoreboard is where 'did it work?' gets a definitive yes or no.
Most scoreboards work by prediction and comparison: an input monitor feeds the expected behaviour into a reference model (the predictor), an output monitor feeds the actual results in, and the scoreboard matches them — often out of order, since real designs reorder, pipeline, and buffer. Getting this matching right (by tag, by address, by stream) is subtle, and a sloppy scoreboard either misses real bugs or screams false alarms. A rigorous, self-checking scoreboard is what lets thousands of random tests run overnight unattended and still mean something.
A scoreboard that never fires a mismatch is suspicious, not reassuring — verification engineers deliberately inject errors to prove the scoreboard can actually catch them.