Analysis: Party Matchmaking Is Broken and How To Fix It
SUMMARY
At lower ranks, matchmaking is biased towards solo and duo queue players. Data from 158 matches shows a clear bias. I assume the opposite is true in higher ranks, but that exceeds the scope of this analysis.
METHODOLOGY
I pulled my last 158 games from statlocker. I used statlocker because it quickly displays party size. I ignored games where someone abandoned. I did not analyze more games, because it reached back to when statlocker was having difficulty pulling match data, leading to many incomplete matches.
Track your Deadlock matches, analyze performance, and improve. Real-time stats, hero tier lists, build guides, and rank tracking.
statlocker.gg
RESULTS
Solo: 17/26 wins (65% winrate)
Duo: 42/75 wins (56% winrate)
Trio: 15/46 wins (33% winrate)
4-stack: 3/11 wins (27% winrate)
DISCUSSION
My account is currently ranked in low Archon. I've been as high as Archon 6 when I solo queue, but every time I play with my friends, matchmaking insists I lose. Having a 30% winrate for party queue is not healthy or fun. I've been steadily improving and am outperforming my account's rank (65% winrate) but not ranking up because of losses in a party. My lower rank friends don't even want to queue with me if I'm on a win streak, because all of us feel the massive change in difficulty when we party up.
I had the same issue with Smite. I was a Masters level player and regularly queued casual games with my friends who were much lower rank. The games
felt the same as in Deadlock, where you are likely using the same matchmaking algorithm (or similar). My games queued in party are
exceptionally and
noticeably more challenging.
I think there's an assumption that the more players are in a party, the stronger that party is. This is likely true at the highest level, but it flips on its head when you get to lower than average rank. The alchemist and arcanist games I play in party are massively more challenging than my solo queue archon games. The individual lane opponents are often phantom or oracle players (2 full ranks above my account) while my random teammates have no clue how to play.
SOLUTION
The simplest and cleanest solution is to have unique matchmaking for individual parties. When I play regularly with my 3-stack and 4-stack, the game should assign a unique rank to the party, not individual players. It should remember this rank and apply it every time we queue together. Wins and losses should
not apply to our individual accounts/ranks.
How this would look:
I queue with my friends for the first time. Our ranks are Initiate, Seeker, Ritualist, and Archon. The game averages our rank to treat us as four fresh accounts ranked somewhere around Arcanist. Conversely, treat us as 4 solo queue players for the first match matchmaking.
After that match finishes, adjust that Arcanist rank based on the win/loss and then treat that as our rank going forward. So next weekend we queue together, and the algorithm treats us as 4 arcanist players.
As we win/lose, our collective rank will rise/lower to meet an approximate 50% winrate. At this point, matchmaking will have done its job and found balance. Even if I climb to Eternus or fall to Initiate, matchmaking will strive to keep our 4 stack at 50% winrate, as opposed to the current 27% winrate. This will create a better experience for all 12 players in the lobby.
KNOWN PROBLEMS WITH MY METHODOLOGY/CLARIFICATIONS
158 games is a small sample size. 11 games as a 4-stack is an even smaller sample size. However, my group has been feeling this problem for the past year.
Individual characters have individual hidden MMR. I did not account for that in my analysis. However, most of my games with lower MMR characters have been losses in 4-stack or 3-stack games, because the current algorithm strives to put me on lower individual MMR characters when I queue into lower average MMR lobbies. The result is that when I'm in a 4-stack, I'm on a weaker character, and as I result I have even less impact, making the match that much harder to carry (which the algorithm expects of me as the highest rank player on my team).
Smurfs exist, and I know some of my losses were to smurfs. When I 4-stack I'm less likely to get a smurf on my side, though (2 possible teammates as smurfs versus 5 if I solo queue).
Many of these games were during the holiday challenge. I'm sure many players were not on their main characters.
THAT'S IT. PLEASE FIX SO MY FRIENDS CAN ENJOY PLAYING THE GAME WITH ME AGAIN.