Innovations in Online Gaming: LoL and its MMR

Matchmaking in League of Legends is the system that decides who you play with and against based on something called MMR (Matchmaking Rating), which is a hidden number Riot uses to measure your skill level and create games where both teams have roughly a 50% chance to win.

The Real Problem With Old Matchmaking

Here’s what was happening: players would win a game and get +14 LP, then lose one and drop -22 LP. That math doesn’t work. You could have a positive winrate and still fall in rank. Riot called these “negative LP states” and they were killing player motivation across every server. And it wasn’t just one region complaining. Iron through Silver players in China had LP gaps of 7 divisions within the same match. Seven. That’s like putting a Bronze 4 player against a Silver 1 and calling it fair.

What Riot Actually Fixed in 2024

Riot reduced queue times for Iron through Silver players by about a minute and brought LP gaps between teams down to around 1 division. The China server saw the biggest improvement where LP gaps dropped from 7 divisions to 1 and average queue times fell by 30 seconds.

The negative LP problem? Players should no longer see themselves enter negative LP states when climbing. That nightmare scenario where you needed a 65% winrate just to break even is mostly gone now. Riot also eliminated instances with 2:0 autofill disparity or worse, which means you won’t face a team with two autofilled players while your team has none.

Another change that flew under the radar: new accounts now need 10 normal games before they can play ranked. This helps the system figure out where someone actually belongs before throwing them into competitive matches.

TrueSkill2: The Microsoft System Riot Is Testing

This is where it gets interesting. Riot is evaluating a system called TrueSkill2 which can better help place smurfs. TrueSkill2 isn’t just some minor tweak. Microsoft developed it and it predicts match outcomes with 68% accuracy compared to 52% for the original TrueSkill.

What makes TrueSkill2 different from the old system? It uses Bayesian inference (fancy statistics that updates predictions based on new data) and factors in things the original never considered: player experience, squad membership, kill counts, and whether someone has a habit of quitting games. TrueSkill2 incorporates player experience, membership in a squad, the number of kills scored, tendency to quit, and skill in other game modes.

The core idea behind both systems is measuring uncertainty. When you have only played 5 games, the system doesn’t really know your skill level yet. In TrueSkill, rating is a Gaussian distribution with both a skill average and a confidence value. The more games you play, the more confident the system becomes about where you belong. The primary purpose is to minimize the number of games necessary to find out a gamer’s skill in order to optimize matchmaking.

If TrueSkill2 works for League, smurf detection would become way more accurate. Right now, a Diamond player on a new account can stomp through low elo for dozens of games before the system catches on. TrueSkill2 could cut that window significantly by looking at how someone plays, not just whether they win.

The Smurf Problem (And Why It Won’t Die)

Smurf accounts for higher skill players still take too long to get to their true rating. Riot knows this and they’re working on it. The 10 normal game requirement before ranked helped seed new accounts better. Better placements with the 10 normal game requirement help seed new accounts into ranks where they average 50% win rates. But experienced players creating new accounts still climb slowly, and every game they spend in Bronze or Silver is a bad experience for the actual Bronze and Silver players they’re destroying.

Beginning with patch 25.18, Riot implemented detection safeguards aimed at identifying smurf and booster accounts with high accuracy. The company stressed that measures are designed to punish offenders without wrongly affecting innocent players.

The 50% Winrate Conspiracy (Debunked)

Some players believe Riot forces everyone toward a 50% winrate by giving them bad teammates after win streaks. Riot’s Lead Gameplay Designer stated this was false, explaining they only match people based on their skill and want to make sure every match feels fair, meaning each team has an equal chance to win.

The system aims for 50% win probability per game, not for your overall account. Perceived streaks happen naturally because matchmaking can’t perfectly predict outcomes. Riot emphasized matches are not manipulated to enforce specific outcomes but instead designed so both teams enter with an equal chance of winning.

Cross Queue MMR: The Next Frontier

One issue still being worked on: when high ranked players play normal or flex games occasionally, it can cause quite imbalanced matches as their solo queue rating isn’t referenced. Imagine a Challenger player hopping into a normal game. The system might put them against Gold players because their normal game rating is much lower than their ranked rating.

Riot wants to fix this by considering overall skill level across queues for matchmaking purposes. It’s complicated because some people genuinely play differently in ranked versus normals, but right now the separation creates lopsided games that frustrate everyone involved.

How MMR Actually Works Behind The Scenes

Your LoL MMR determines everything: who you play against, how much LP you gain or lose, and how fast you climb. The visible rank is just a badge. MMR is the real number.

When you beat higher MMR opponents, you gain more. When you lose to lower MMR opponents, you lose more. The system constantly adjusts, trying to find matches where both teams sit around that 50% win probability. Matchmaking attempts to create fair matches where both teams have a truly equal probability of winning.

Importantly, statistics such as scorelines including KDA and creep score are not considered since the system would be open to exploiting. If KDA affected MMR, players would play selfishly to protect their stats instead of making risky plays that help the team win. Win or lose is the only thing that matters for rating changes.

What This Means For Your Climb

The matchmaking improvements won’t magically boost your rank. They just make the system fairer. You still need to actually improve at the game to climb. But at least now you won’t feel like you’re fighting the system itself alongside your opponents.

Queue times dropped. LP disparity dropped. Autofill nightmares dropped. And if TrueSkill2 gets implemented, smurf detection should improve too. None of these changes make climbing easier, but they make climbing feel less arbitrary. And sometimes that’s exactly what players need to stay motivated through the grind.