Does your last game predict your next one?
"Loser's queue." "I'm tilted, one more." "Never sleep on a loss." "Momentum is real." We lined up the consecutive ranked games of tens of thousands of players and measured what actually happens after a win, after a loss, after a losing streak, after a stomp, after a bad personal game, and depending on how fast you queue again.
- Loading…
How to read the two numbers
Every condition gets two results. The raw win rate is simply how often the next game was won. But players who lose a lot are on losing streaks a lot, so raw numbers mostly measure who is on a streak, not what the streak does. The second number, vs own baseline, compares each game with that same player's win rate over all their other games in the sample: it answers whether the condition itself moves the odds. It is shown in percentage points (pp); 0 means "no different from your usual".
- Sample: a random 90,000 players with match histories, every competitive game of theirs over roughly 60 days, results fetched for all of them. A game counts when its own result and the previous ranked game's result are both known.
- A session ends after a 6-hour gap. Streaks and session records are counted inside a session; "sleep on a loss" looks at the first game after such a gap.
- Match matchmaking works against streaks: after losses your hidden rating dips and you get slightly easier opponents, after wins the reverse. The baseline-adjusted number includes that effect — it is what you actually experience.
After a win vs after a loss
Streaks: is there a loser's queue?
Win rate of the next game by the streak you are on inside the session. If a loser's queue existed, the bars on the left would sink below your baseline the longer the streak; if matchmaking simply regresses you to the mean, they rise.
How you lost (or won) the last one
A 13-2 stomp and an overtime heartbreaker are both "a loss". Do they leave the same mark on the next game?
Requeue timing: "one more" vs a break
Time between the end of the previous game and the start of the next, split by whether the previous game was a loss or a win.
Sleeping on a loss, and long sessions
Across ranks
Caveats
- Baselines are estimated from the same sample. Each player's baseline leaves the game in question out, but with 10–30 games per player it is still noisy; the cluster-robust intervals account for that.
- Who requeues fast is not random. A player who instantly queues after a loss may be the calm type or the tilted type; the timing results describe what happens, not what would happen if the other type did it.
- Only ranked games are counted for streaks. A deathmatch between two ranked losses does not reset anything here.
- Surrenders and early leaves are bucketed separately when the round count reveals them; a player who left mid-game still counts as a loss.