You do not lose rank in games, you lose it in sessions

Pull up almost any stuck account’s ELO graph and it has the same shape three times over: climb about 300, hit a new peak, give back 400. Repeat.

That shape is not variance. Variance is noise around a trend. This is a pattern, and patterns have causes.

Group your matches into sessions

Take your match history and split it wherever there is a gap of more than about three hours. Each group is one sitting. Then look at net ELO per session rather than per match.

What almost everyone finds:

• Short sessions are roughly break-even or positive.
• Long sessions are heavily negative.
• The worst sessions are late at night.
• The worst of the worst continued after two straight losses.

A single bad match costs 25 ELO. A bad session costs 150–250. Six bad sessions undo two months. That is the entire drawdown, and it happens on maybe eight nights a year.

Why more matches makes it worse

Three things stack up inside a long session, and they compound rather than add.

Attention decays. By match six you are not making worse decisions because you are upset — you are making worse decisions because you are tired. Crosshair placement is the first thing to go, then utility (you stop buying it), then comms.

Losses change how you play. After two losses most players get faster. Faster peeks, earlier duels, more solo entries. Precisely the behaviours that lose more rounds, adopted at the moment you can least afford them.

Sunk cost takes over. “One more to get it back” is the sentence at the centre of every drawdown. It has never once been true, because the ELO you are chasing was lost by the state you are currently in.

The two rules

Rule one: two straight losses ends the session.

Not a break. Not “one more”. The session is over. This is uncomfortable precisely because it fires at the moment you least want it to — that is what makes it work.

Rule two: a new peak ELO also ends the session.

This one is counterintuitive and it is the more valuable of the two. You just hit your highest ever rating. The urge is to keep going. But you are now playing with the most to lose and, typically, several matches of fatigue already spent. Bank it. Log off. The peak is only real if you stop.

Two supporting rules

Cap the session at five matches even when you are winning. Fatigue does not check the scoreboard.

Do not queue after 23:00. Look at your own late sessions before dismissing this — for most players the late-night win rate is several points below their daytime rate, and it is where the drawdowns cluster. If you can only play late, then the five-match cap and the two-loss rule matter twice as much.

Why willpower will not do it

You cannot decide, in the moment, whether you are tilted. Tilt’s defining feature is that it feels like clarity — you feel focused, you feel like you are about to turn it around.

Which is why both rules are mechanical. Two losses. New peak. No judgement call, no self-assessment, no negotiating with yourself at 1am. You either follow the rule or you do not, and you can check afterwards whether you did.

Make it visible

Keep the record for thirty days. Date, matches played, net ELO, and two yes/no columns: did I stop at two losses, did I stop at the peak.

Thirty days is enough to see your own pattern, and seeing it is most of the fix. People who dispute that tilt costs them anything usually stop disputing it after two weeks of their own data.

Then compare the ELO you actually gained against what you would have gained by simply not playing your five worst sessions. That number is normally larger than anything a mechanical improvement would have produced in the same month.

See your own session pattern — grouped, scored, with the two rules checked for you.