A league going quiet rarely feels like a decision. Nobody announces they're leaving. The Thursday night just gets smaller, then it gets cancelled once, then twice, and three months later there's a leaderboard nobody has looked at since spring.
The usual explanation is that people got bored, or a new game came out, and sometimes that's all it is. But leagues have a very consistent shape to how they empty out — they hollow from the top down, not evenly — and once you know the shape you can see that part of it is something your league did to itself.
Some of it is the rating system. Not because it's broken; because it's working exactly as designed.
The Best Player in Your League Has a Terrible Bet
Elo updates a prediction. When a heavily favoured player wins, that outcome was expected, so it teaches the system almost nothing and the rating barely moves. When they lose, that's genuinely surprising and it moves a lot.
That's correct behaviour. It's also, from the strongest player's point of view, an awful proposition. Here's what a player is actually risking each time they queue against a 1200-rated field, at K=32:
| Your rating | Opponent | Win | Loss | Downside ratio |
|---|---|---|---|---|
| 1200 | 1200 | +16.0 | −16.0 | 1× |
| 1400 | 1200 | +7.7 | −24.3 | 3× |
| 1600 | 1200 | +2.9 | −29.1 | 10× |
| 1800 | 1200 | +2.9 | −29.1 | 10× |
The person at the top of your ladder is playing a game where a win is worth about three points and a loss costs about thirty. They have to win ten in a row to undo one bad night.
Nobody experiences that as arithmetic. They experience it as there's nothing in this for me, and they queue a bit less. That's the first crack, and it opens at the top.
What the Caps Are Preventing
It's worth knowing how much worse this could be, because the defaults are doing real work here.
Rating caps — on by default — clamp the effective rating difference at 400 points, so a huge gap is calculated as though it were merely a large one. Without them:
| Gap | Capped win | Uncapped win | Uncapped downside ratio |
|---|---|---|---|
| 400 | +2.91 | +2.91 | 10× |
| 600 | +2.91 | +0.98 | 32× |
| 800 | +2.91 | +0.32 | 100× |
| 1000 | +2.91 | +0.10 | 316× |
Uncapped, a dominant player in a small pool ends up risking three hundred times what they stand to gain. The caps hold the ratio at 10× however far ahead someone gets, which is the difference between a bad bet and an absurd one.
But read that honestly: 10× is still a bad bet. The caps stop the problem escalating; they don't remove it. If your league has one clearly best player and no peers for them, no rating configuration will make queueing attractive to that person. That has to be fixed somewhere other than in the maths.
Why It's a Spiral and Not a Dropout
Here's the part that turns one person's disengagement into a dead league.
When the top player stops queueing, the ceiling of your pool drops — and whoever is now at the top inherits exactly the same bet:
| Ceiling player | Field average | Win | Loss | Downside |
|---|---|---|---|---|
| 1600 | 1200 | +2.9 | −29.1 | 10× |
| 1450 | 1100 | +3.8 | −28.2 | 7× |
| 1300 | 1000 | +4.8 | −27.2 | 6× |
The structure regenerates itself one tier down. Each departure re-creates the conditions that caused it, which is why the decline accelerates rather than levelling off, and why "we lost a couple of good players" turns into an empty queue faster than anyone expects.
Meanwhile the bottom of the ladder is leaving too, for the ordinary reason — they lose most games. So the league empties from both ends and the middle goes last, which is why the final stage feels sudden: by the time the middle notices, there's nobody above or below them to play.
Fixing the Incentive
Since the arithmetic can't be argued with, the fix has to give the top of your ladder a reason to play that isn't rating.
Recruit upward, deliberately. The only real solution to "there's nobody who can beat me" is somebody who can. One player above your current ceiling fixes more than any setting, and it's worth treating as a specific recruiting goal rather than hoping the general intake produces one.
Run formats where strong players can affect the outcome. In a 1v1 ladder a dominant player has nothing to prove. On a team, they can carry — and carrying is fun in a way that beating a much weaker opponent isn't. Mixed-skill team formats with balanced teams keep your best players engaged far better than a solo ladder does, because the challenge is real again.
Give them something that isn't a number. Tournaments where winning is an event. A champion title that expires and has to be re-won. A seeded bracket where the top seed is a status worth defending. All of these attach meaning to a win that Elo structurally cannot, because Elo's whole job is to stop being impressed once it expects you to win.
Consider a handicap event, occasionally. Not on the main ladder — as a one-off format where the top players start behind. It's a gimmick and it works, because it restores stakes for the people who don't have any.
Rank Decay Fixes the Board, Not the Problem
Rank decay — ratings drifting down after a period of inactivity — is the feature people reach for here, and it's worth being clear about what it does.
It makes the leaderboard reflect current activity, so the top of your board isn't occupied by someone who stopped playing in March. That's genuinely valuable, and it's fair to the people still turning up: a ladder where an absent player permanently holds first place is demoralising for everyone below them.
What it doesn't do is make queueing rewarding. It's a stick applied to absence, and the person it's aimed at has already decided the league isn't worth their evening — a slowly falling number is unlikely to change that, and it can read as petty if it's the only lever you pull. Use it to keep the board honest, and fix the incentive separately. The tier roles and rank decay guide covers the mechanics and sensible thresholds.
The same goes for minimum-match requirements. Hiding players below a match threshold makes the visible board more meaningful and stops a 3-0 newcomer sitting at the top — worth doing. But it's display hygiene, not retention.
Restarting: You Have Something a New League Doesn't
If it's already quiet, the good news is that a restart is a materially easier problem than a cold start. You have a list of people who already demonstrated they liked this — that's an enormous asset and most restart attempts waste it.
What works:
Pick a date and personally ping people. Not an @everyone announcement in a channel nobody opens any more. Individual messages to the twenty people who played most. A quiet league's members haven't rejected you; they've just stopped having a reason to open Discord on a Thursday, and a personal message from a human is a reason.
Overbook it. The arithmetic in the cold-start post applies directly — you need roughly 50% more yeses than seats, because the show rate on a restart night is worse than usual.
Make the first night an event, not a queue re-opening. A tournament, a draft, a themed one-off. "The queue is open again" asks people to go first; an event with a time gives them something to turn up to.
Start a new season. This is the highest-leverage single move available, because it converts "you're 400 points behind and it's hopeless" into "everyone starts roughly level on the same date". Running a competitive season covers the carryover choices — for a restart, compressing ratings toward the middle is usually right.
Don't apologise for the gap. A "sorry we died, we're trying again" message frames the thing as a failure being retried. A date and a format frames it as a thing that's happening.
When To Let It Go
Worth saying plainly, because pouring effort into a league that's finished is its own cost.
If your active core is under about ten people and none of them are asking when the next night is, the honest read is usually that the game moved on rather than the league failed. Two options beat a slow decline: merge into a larger community running the same game — your members get matches, which is what they wanted — or archive it cleanly, post the final standings, and leave it as something that happened rather than something that fizzled.
A league that ended with a champion is a much better memory than one that stopped mid-season, and it's much easier to restart later.
Setting It Up
- Check your rating spread. If your top player is 400+ above the field, they're on a 10× losing bet and you have a structural retention problem, not a motivation one.
- Leave rating caps on. They bound the asymmetry at 10× instead of letting it reach 300×.
- Recruit upward — treat "someone who can beat our best player" as a specific goal.
- Run team formats with balanced teams so strong players can carry.
- Add non-rating rewards — tournaments, defended titles, seeded events.
- Turn on rank decay to keep the board honest, and don't expect it to bring anyone back.
- Restart with a date, personal pings, an event and a new season — in that order.
Frequently Asked Questions
Should I lower the K-factor to make ratings more stable?
It won't help with this. K scales both sides of the bet equally, so the 10× ratio is unchanged — the top player just gains 1.5 and loses 15 instead. The asymmetry comes from the rating gap, not from K. Lower K for noisy formats, not for retention. See K-factor curves.
My best player says the ladder is pointless. Are they right?
About the ladder specifically, in a small pool, largely yes — and telling them the maths agrees with them tends to go better than telling them to keep queueing. What's actually broken is that your league only offers one kind of reward. Add an event they can win and the complaint usually goes away.
Is rank decay going to annoy people?
If it's the only thing you do, yes. It reads as punishing people for having lives. Announce it in advance, set a generous threshold, and pair it with something people gain by returning — a new season, an event — so returning has an upside rather than just avoiding a penalty.
How long should I wait before calling it?
Longer than one bad month, shorter than a year. The signal isn't the match count, it's whether anyone besides you is asking when the next one is. If two or three people are still asking, you have a league that needs a restart. If nobody is, you have a decision to make.
Can I just reset everyone's rating to bring people back?
A reset on its own changes numbers, not behaviour, and people who weren't playing won't notice it happened. It works when it's part of a season with a date, an announcement and an end — the reset isn't the draw, the fresh start with a finish line is.
Keeping a league alive? Team Up supports rating caps that bound runaway gaps, balanced-team matchmaking so strong players can carry, rank decay and inactivity thresholds to keep the board honest, and seasons with configurable carryover for a clean restart. Start with running a competitive season, or the matchmaking queue guide to get a Thursday night running again.