Taking a character's best feats might be an objective way to handle things, but it's a really really really bad way to answer the core question of a DB: "who would win?" Mostly because it powercreeps characters to no end.

Quick thought experiment: let's say there's a character called "1% Woman". Her power is that 1% of the time, she wins a fight against her opponent, no matter who that opponent is. 1% of the time, she loses a fight against her opponent, no matter who that opponent is. Over the course of the series, we see her take on a wide variety of opponents, and the writers slavishly follow the 1% rule: 1% of the times she fights opponents, she beats them outright and conclusively using this power, even when she has no right to.

And in DB, those are the only times that matter. Let's say she's up against Saitama, good ol' One Punch Man himself. Within the story, what percentage of a chance does she have of beating him? 1%. Within Death Battle, what percentage of a chance does she have of beating him? 100%, because we're cherrypicking her best feats, and there was that one time her 1% power gave her the win over a literal god.

Every character is like this, every single one. Raw power fluctuates, expertise varies, and luck happens. Sometimes you're on a good day, sometimes not so good, sometimes you pull off a feat you didn't realize you were capable of, and you never manage it again.

Interestingly, Deadliest Warrior (for its historical flaws) gets this right. Instead of trying to nebulously find the high point of their combatants, the ideal situations, they build out data based on the weapons and techniques of the warriors, and then feed it into a computer simulation and run it 1,000 times, and the side that won more times wins...and then the hosts talk about the results, and whether they agree or disagree.

Does it produce a definitive, conclusive result? Not at all, because trying to explore a situation where one character beats another 100% of the time, barring clear stomps, is a fool's errand. Death Battle's results might be conclusive, but they're not always true to the characters, and if you ask me, being true to the characters is kinda important.