Mario Kart Shows Why the Fastest Driver Is Not the Best

Before a race starts in Mario Kart, Nintendo's long-running cartoon racing game, you pick a driver. The game shows you their statistics as small bars: how fast the character goes flat out, how quickly they get back up to that speed after a crash. There are dozens of characters. Most players choose the one they like. The competitive question underneath is more interesting: of all those options, how many can be ruled out by argument alone, before anybody's preferences enter the room?

That is the question a recent interactive explainer uses the game to answer. It is nominally about karts. It is really about a piece of nineteenth-century economics that has quietly become the standard tool for choosing among options when you want more than one thing at once.

The speed column is not the answer

If speed were the only thing that mattered, the problem would be trivial. Sort the characters by top speed, take the one at the top, and you would be done in a second. Bowser and Wario sit up there, heavy and quick, and the whole exercise would collapse into reading one number.

But speed is not the only thing that matters. Acceleration matters too — every time you hit a wall, take a shell, or fall off the track, you have to climb back to full speed, and how long that takes affects whether you finish first. The two attributes trade against each other: the heavy characters that go fastest recover slowest, and the light ones that recover instantly never reach the same top end.

The moment you are looking at two numbers instead of one, sorting stops working. There is no single column to rank by. And this is the point at which most people conclude that the choice is purely a matter of taste — that with two things to want and no way to have both, you may as well pick your favourite.

Some characters are simply worse

They would be wrong, and the reason is visible if you plot the characters on a chart with speed on one axis and acceleration on the other.

Take the humble Koopa. There is another character, Cat Peach, with exactly the same acceleration and more speed. There is a third, Toadette, with exactly the same speed and more acceleration. Koopa is not making a trade-off. Koopa is not offering you anything in exchange for anything. Whatever you happen to value, one of those two characters gives you more of it and no less of the other.

The technical term is that Koopa is dominated. And the useful property of domination is that it does not depend on what you want. A player who cares only about top speed should not pick Koopa. A player who cares only about recovery should not pick Koopa. A player weighting the two in any proportion at all — any proportion, no exceptions — is better served by something else. You can delete Koopa from the list without knowing a single thing about the person choosing.

Do that for every character and the list shrinks dramatically. What remains is the set of options that nothing else beats outright: each one is the fastest available at its level of acceleration, or the most responsive available at its speed. Drawn on the chart, they trace an outer edge. That edge is the Pareto front, named after Vilfredo Pareto, the Italian economist who formalised the idea around the turn of the twentieth century. The same construction now underpins the whole field of multi-objective optimisation — the branch of applied maths concerned with problems that have more than one thing to maximise.

What survives on that frontier is not a winner. It is a shortlist, arrived at without anyone expressing an opinion.

The frontier tells you where to look, not where to stand

Being on the frontier does not make options equal. They are spread along it, from the extreme speed end to the extreme acceleration end, and the explainer suggests most players will want something in the middle rather than either edge. Which point suits you depends on how you actually play: a driver who rarely crashes has little use for fast recovery, and a driver who crashes often has little use for a top speed they never hold.

That is where the objectivity runs out. Pareto efficiency can tell you which choices are indefensible. It cannot tell you which of the defensible ones is yours.

What the people who race for records actually pick

The suggestion that you will end up in the middle is where practitioners push back hardest.

People who set speedrun records in the series report going to the extreme instead — the heaviest, fastest characters, at the very corner of the frontier, in both the original game and the modern entries. Their reasoning is that acceleration is insurance against mistakes, and at their skill level they do not make enough mistakes to need it. Players who study competitive builds go further and argue the explainer has picked the wrong second axis altogether: in serious play, acceleration is close to irrelevant, and the real trade-off is between speed and mini-turbo, the boost you earn by drifting through corners.

Others raise a sharper objection to the method itself. The whole construction assumes more of a thing is always better. If too much speed sends you off the track, the value of speed peaks somewhere and then falls, and the outer edge of the chart is no longer where you want to be.

None of that damages what the frontier does. It is worth noting how far the pruning scales: someone optimising character builds in an online role-playing game with fifteen equipment slots and hundreds of items per slot — more combinations than atoms in most analogies — pruned dominated items slot by slot, merged the survivors in pairs, and pruned again, reducing an intractable search to a couple of seconds of computation. Approaches that skipped the pruning ran for minutes, or effectively forever.

The frontier is the same instrument at both scales. It settles what can be settled and hands you back a much shorter list, along with the news that the rest is up to you.