← Insights

Does switching character after losing a game actually help?

Measured over 1,572,427 games · 19 August 2026

You drop game one. The rules hand you a counterpick, and with it the temptation to switch character. Everyone has an opinion about whether that works. Nobody seems to have counted.

We have the full game-by-game record of 870,660 Smash Ultimate sets, including which character each player used in each individual game. So we counted: after losing a game, how often does that player win the next one — and does switching change it?

The answer depends entirely on who you are

Pooled together, switching looks bad: 34.6% versus 38.3%. But pooling hides the real finding, because underdogs both switch more often and lose more often. Splitting by the rating gap — comparing underdogs with underdogs — the pattern reverses partway down.

After losing a game, how often that player wins the next one
Their rating − opponent'sKept characterSwitchedDifference
below −4004.7% (151,357)3% (29,295)-1.7pp
−400 to −30012.7% (106,036)8.7% (16,818)-4.0pp
−300 to −20020.6% (168,966)15.2% (22,840)-5.4pp
−200 to −10031% (242,818)25% (28,027)-6.0pp
−100 to 043.1% (275,974)38.4% (29,734)-4.7pp
0 to +10056% (224,897)54.3% (25,149)-1.7pp
+100 to +20067.3% (129,735)68.3% (16,711)+1.0pp
+200 to +30076.7% (55,498)79.2% (9,871)+2.5pp
+300 to +40083.7% (20,246)86.8% (4,931)+3.1pp
above +40090.4% (9,723)94% (3,467)+3.6pp

The gradient is clean and it crosses over between a 100 and 200 point rating advantage. If you are clearly the weaker player, switching costs you one and a half to six percentage points. If you are clearly the stronger player, it gains you one to four.

Why the strong get paid and the weak do not

A separate measurement on the same data explains it. We fit every player's rating on each character they play — how strong Glutonny is as Link, not how strong the average Link is. Almost everyone's secondary comes out weaker than their main, often by a hundred points or more.

So for most players, switching means moving to a character they are simply worse on, in the hope that the matchup makes up the difference. It usually does not. The players it works for are the ones with a genuine second character — deep enough that the switch is adaptation rather than a downgrade.

What this is not

This is an association, not advice. Players choose when to switch, and that choice carries information we cannot see: how one-sided the previous game was, what the stage was, whether they had a read. Someone who gets three-stocked and switches was probably losing that set either way, and we cannot separate the switch from the situation that prompted it. Nothing here says "do not switch" — it says players who switch while behind go on to win less often than players who do not.

Rating gaps come from our Whole-History Ratings, which are recomputed weekly over every set from start.gg. Games are restricted to sets where the complete sequence was reported, which is 99.9% of sets that have any game detail at all.

Beta Whole-History Ratings, recomputed weekly from start.gg brackets — every past result revalued as opponents' true strength emerges. How the rating works