The Patents & Systems in League of Legends Rigged Matchmaking ~ Engagement Optimized Matchmaking

Topic and purpose

• Topic: An in-depth look at League of Legends matchmaking through the lens of engagement-optimized systems, behavioral-matchmaking patents, machine-learning churn prediction, and how those systems could make ranked matches feel “rigged.”
• Purpose: Explain research, patents, and developer presentations that connect matchmaking, player behavior scoring, and retention/monetization goals — and give practical tips (some superstitious) for players who think they’re being routed into “loser” or “toxic” queues.

Main concepts and claims

Engagement-optimized matchmaking

• Originates from academic and industry work (example: EA + Northeastern, 2017). The core idea: an engine can nudge player retention by controlling match outcomes and the player’s final impression (the peak–end rule).
• Machine learning (ML) can predict churn risk and change the experience to keep players playing — e.g., mixing easy and hard matches to maximize continued play.

Churn and the peak–end rule

Churn = not returning for 7+ days.

• Players’ memories of the final match of a session heavily influence whether they return; systems can exploit this by engineering a “good final impression.”

ELO / MMR / visible LP system

• League uses a hidden numeric matchmaking rating (MMR/ELO) behind the visible rank + LP system.
• Because MMR is hidden, matches can feel opaque and players cannot verify whether match balance is fair.

Behavioral-based matchmaking (Riot patent)

• Riot’s patent — “Systems and methods that enable player matching for multiplayer online games” — describes matching players using behavioral data from profiles: honors, reports, chat behavior, playstyle metrics.
• Typical patent concept:
  • Assemble a candidate pool by skill.
  • Use behavior/personality vectors to place players on teams (e.g., “leaders”, “teammates”, “verbal aggressors”).
  • Store report/commend data and compute personality/behavior scores.
• Consequence: players with many reports (or many honors) could be routed into different queues (video terms used: “reported Q”, “loser Q”, “toxic Q”).

Rubber-banding and historical context

• Classic arcade machines and some fighting games used difficulty-scaling / “cheating AI” to extract more credits — a historical metaphor for modern engagement optimization.
• Modern systems are ML-driven and subtler, but the intent — maximize retention/revenue — can be similar.

Machine learning and user-engagement cycles

• Industry presentations (e.g., Activision/AWS) show ML mapping player “states” and sequencing actions/messages to steer players toward high-engagement states and away from churn.
• Systems can predict churn days or weeks ahead and select interventions (easy match, loot, restrict autofill) tailored to players’ predicted behavior.

Real-time modification — the deepest concern (speculative)

• An Activision patent describes dynamically modifying in-game content in real time based on external conditions or desired player experience (examples from Call of Duty).
• The video raises the question: could League (or other live games) modify in-game variables (damage, hit registration, survivability) to engineer an experience? This is presented as plausible given patents and precedent, but remains speculative for League specifically.

Observed matchmaking patterns & why players distrust it

• Frequent, extreme streaks: long win streaks followed by long loss streaks reported more often than random even-skill matches would produce.
• Mixed-rank lobbies: Bronze through Plat in the same match and many one-sided stomps.
• Disruptors: smurfs, boosters, bot farms, griefers, and derank streams make balance difficult.
• Low player agency: power-creep and certain champion designs let single players (especially feeders) decide many games, increasing perceived unfairness.
• Match history/labels: historical labels (e.g., “hot streak”, “fresh blood”, “veteran”) and the push to rapidly place players (TrueSkill2) can conflict with retention incentives.

Game-design role in perceived manipulation

• Champions and power creep: many modern champions are snowbally and highly mobile; a single fed player can end games quickly, increasing variance and emotional highs/lows.
• Mechanics that reduce player agency (vision changes, role-quests, strong mobility) can make matches feel more like chaotic outcomes than pure skill contests.
• Design choices combined with behavior-based routing amplify the emotional rollercoaster that maximizes engagement.

Practical tips, strategies, and player “folk remedies”

General play / skill advice

• Prefer snowball/carry champions if you want to climb (examples: Darius-, Briar-, Viego-style champions).
• Maintain solid performance stats (KDA, CS, vision, objective participation). These metrics feed behavior/skill profiles.
• If you want to avoid being marked as an outlier smurf, don’t always post huge stat lines — sometimes “blend in” for more consistent matches.

Behavioral / social-credit style tactics (avoid “reported/toxic” queues)

• Honor teammates regularly and encourage positive comms; honors and reports are tracked.
• Avoid excessive or frivolous reporting—over-reporting may hurt perception or behavior signals in some systems.
• Report only clear verbal abuse or targeted toxicity; mass-reporting feeders can backfire.
• Avoid sarcasm, toxic chat, or attention-seeking behavior; text/voice analysis may penalize nuance.

If you’re in a bad streak / “reported Q”

• Stop queuing: some recommend not playing for a full week (7+ days) to reset churn/behavior labeling and get easier matches on return.
• Use normals to earn honors and reset behavior signals before returning to ranked.
• Submit a support ticket if you believe you’re mis-scored — human review can sometimes change behavioral flags (not guaranteed).
• Preserve personal stats in bad matches: AFK-farm safely rather than producing extremely poor KDAs that might hurt your profile.
• Occasional leaving: leaving once in a while has light penalties and might break a negative cycle, but repeated leaver behavior risks punishment.

Queue timing and other superstitious tips

• Instant queues: considered a red flag by some — instant matching can mean the game filled a slot and the match may be worse.
• Long queue times: sometimes correlate with better matches (system searches for compatible, non-toxic players).
• Expect “skill-check” matches when hidden MMR outruns visible rank — harder games may be intended to slow a climb.
• If tilted, quit for the day — don’t chase losses.

Other operational tips

• Be mindful of autofill/hot-streak protections: the game gates autofill under certain conditions to protect streaks.
• Avoid drawing attention in chat; public recognition (good or bad) changes behavior vectors.
• If you genuinely believe you were unjustly reported or mis-scored, support tickets are the route to human correction.

What is likely vs. speculative

Likely / supported

• Riot uses behavioral signals in some capacity (patent and patch-note references to behavioral systems).
• ML and churn prediction are standard industry practice and likely used at scale in big live games.
• Smurfs, boosters, griefers, and bot farms materially affect match quality and are difficult to fully prevent.

Speculative / unproven

• Real-time dynamic modification of damage/hit registration in League specifically to craft match outcomes is plausible (given industry patents and examples) but not proven for League.
• Exact implementation details of Riot’s TrueSkill2, internal weights, and how labels influence every match remain proprietary and opaque.

Takeaway / final perspective

• Big live games have both the incentives and the technology to steer player experiences toward maximal engagement. Riot has a patent describing behavioral matching, and Riot’s systems (TrueSkill2, behavior systems, honor/reports, hidden MMR) create plausible mechanisms for players’ repeated experiences of extreme streaks and apparent unfair matches.
• Even if not all aspects are confirmed, the combination of lack of transparency, industry precedent, and documented patents makes it reasonable to suspect retention-driven components in matchmaking.
• Practical response for players: protect your behavior stats, avoid chasing losses, play safely when a match is hopeless, be strategic about when and how you queue, and consider extended breaks to reset behavior-related signals.

Sources and people featured (as cited in the video)

• Josh Mink (matchmaking designer; presentations referenced).
• Riot Games — patent: “Systems and methods that enable player matching for multiplayer online games.”
• EA + Northeastern University (2017) — engagement-optimized matchmaking / churn prediction paper.
• Activision / AWS presentation (2019) — ML for player states and engagement.
• Activision patent: “systems and methods for dynamically modifying video game content…” (real-time modification examples).
• Professor Elo (ELO system), TrueSkill / TrueSkill2 references.
• Tencent — Riot’s parent company (mentioned in context).
• AlphaGo — cited as an ML benchmark.
• Other examples: arcade rubber-banding, champions (Amumu, Rammus, Briar, Darius, Viego), Vanguard anti-cheat.