[2022] Please fixed ranked riot

Keep in mind that 70/30 is not the most common, it's just the worst possible. The most common is still 50/50 and then it decreases in occurrence from there.

Also, it's not just wait time we are trading off against, it's position preference. Before position preference, most matches were close to 50/50. When you get a less fair match it's almost always because it was trying to make sure everyone got the lane they wanted. Once it has 10 players within the allowed rank restrictions, it will often prioritize getting you your position over optimizing the fairness. If possible it will still aim for close to 50/50, but it's currently "allowed" to go higher to keep positions.

Those are good suggestions and they are the types of things we discuss pretty often. It's a bit tricky to allow a "wait longer" option because if a given player decides to wait longer, then all the other players who may need that player in their match also have to wait longer. Basically, we have to balance very deliberately any feature that allows one player to force another player to wait longer.

Ironically, it's also the higher-ranked players where finding matches becomes the most difficult because, by definition, there are always fewer of them. The current balance has been allowing many players to find matches who, previous to this patch, could not play at all. So making it more picky at the top would currently prevent many of them from being able to play Ranked at all.

That said, we actually do allow up to 20-minute wait times. But like I alluded to above, when it comes to finding as fair a match as we can within 20 minutes, it isn't so much the difference between a 3- or 10-minute wait time, it's the difference between finding a match or not finding one at all, even within 20-minutes. The system has some capability to ask the question "Can I find something better in 10-20 minutes? If so maybe we'll wait longer, but if it's not going to get better, there's no reason to make people wait for the same overall match"

Also, even within fairness, there are two measurements that need to be balanced. There's the team vs. team fairness that we have been discussing (like, 50/50, 60/40, etc.) but there's also "Are my teammates as good as I am?" If we only tighten up the team vs. team fairness, then the system will end up forcing worse teammates onto the teams of good players in order to maintain fairness. This will create matches that are, yes, 50/50, but will feel like "coinflips" because the good player will not have much influence on the win, or will at least have to hard carry a lot of their matches. To fix that, we would have to also tighten the allowed skill gap between all of the 10 players that get into the match. This is actually much harder to do than just team balancing though. You can team balance 10 random players pretty well by putting the same amount of skill on each team. But finding 10 high-skilled players that are all really close in skill together usually requires a very long wait time. Wait time actually grows exponentially as you increase teammate quality at the high end. So it's not just 3 vs. 10 minutes, it's 3 vs. 60 or 120 minutes.

So creating a good match combines team balance, teammate quality, position preference, latency, and wait times. If we tried to super tighten up all of them, then there would have to be a lot players who can't play. So instead, we try and balance all of them within a reasonable wait time.

We are constantly evaluating how much tighter each dimension can be though, so if possible, we will increase the team vs. team fairness, and also the teammate quality, whenever we can.

What rank are you?

The current allowed range is +/- 1 from the anchor player or party's rank, but there are exceptions at the top (Diamond+) because otherwise they wouldn't find matches.

So if you are Platinum, you can match both -1 (Gold) and +1 (Emerald). So if the anchor player is Platinum, you can have both Gold and Emerald players in the match.

If you are Diamond, you can match -1 (Emerald) but you can also match up to Challenger because population thins out super fast GM+, so Challengers can't play w/o Diamonds. So if a Challenger player is the anchor --- which is common because anchors are usually whoever has waited the longest, and Challengers often wait the longest because of high rank and MMR --- then you can have Diamond all the way up to Challenger players in the match. Though you can't have Challengers with Emeralds unless you have, e.g. an Emerald/Diamond/GM trio that averages to Diamond.

If your rank is in those higher levels though, either you, or the ranks above you, would not be able to find matches if we reduced those ranges.

If the anchor is Emerald, they can match -1 (Plat) and +1 (Diamond) so you can see Plat to Diamond in the same match.

A GM anchor can match high Emerald ( I only) to Challenger so you can see Emerald I to Challenger in the same match.

Other combinations require parties since a party's rank is the average of its members.

These are valid concerns.

1. In the data for the last month or so, I don't see Golds ever in matches with Masters. There was a window where we were testing the matchmaking parameters overall and did allow that, but we have since closed that off. Do you know when you saw that? If you have a date and screenshot so I can pull the match, I can look. So far every example I've looked up was legitimately within the rules. The only strange matches that happened involved a trio where 2 higher-ranked players brought in a lower one, stretching the gap.
2. We are aware of how stacking is affecting both fairness and the leaderboard and are actively considering solutions. On the one hand, players in 5q are pretty happy at the moment because ("having the most fun I've ever had in WR!"), which we don't want to take away, and they only match other 5q parties. However, we also know they often have the potential to climb faster than solo queue because of higher win%. We are investigating a number of solutions that range from not allowing higher 5q to play (which would be sad) to slowing their VP climb to be in line with solo queue, which would be messy but perhaps a compromise. We don't have anything to official to say on that though, and your feedback is welcome. For trios, if you have a trio, so does the other team. But, yes, it is possible that one team has a higher-skilled trio and worse solo queue fills, whereas the other side has a lower-skilled trio and high skilled fills. We are investigating this situation as well. It could be the best compromise is to not allow trios, and to slow down the 5qs.

Another solve we are looking into is tightening up the player to player skill gaps within teams. This would make it so when you do have a trio vs. trio, it's less likely there's a large gap between those two trios, and likewise the duos and solos that fill the other 2.

Some data that may be useful to know:

Win% of trios with at least one Challenger or GM: 55%

Win% of 5q with at least one Challenger/GM: 67%

So trios do seem to have an advantage at high rank, but it's not crazy on average. Though, like you said, there are probably some stand out occurrences we could tamp down by tightening up the allowed within-team skill gaps.

Yeah, I totally agree that it is better to have players as close to your skill as possible on your team.

Improving this requires tightening the allowed player to player skill gap, making it more likely that each other player on your team (and in the whole match in fact) is closer to you in skill.

This is something we are actively working on. One trick here is that wait time grows exponentially with this particular gap. So it becomes not a question of, "I'm willing to wait another 10 minutes" but instead another 60-120 minutes (as I mentioned in a previous post). Especially when you need to find 10 players at the high end all close in skill.

To deal with this, the current system detects real-time availability and tries to give you the closest 10 possible players. So sometimes, it can make very tight matches. At other times, the system gives up because it knows that even after 20 minutes, the match quality will be the same. We do have some settings within this system that we are looking into tightening. Another wrinkle here is that when it looks for those 10 possible players, it also has to consider positions preferences. So it's not just "find the 10 closest to me" it's find the 2 closest Junglers to me, then the two closest who prefer mid, etc. So the skill gap between you and each position is related to how popular that position is at your skill level at the time you search.

We are constantly looking at how much tighter we can all of the matchmaking quality related settings, including the player to player skill gap, as well as the team to team one, while also avoiding , e.g. Challengers not finding matches at all. My personal preference (not Riot official) would be to all get everyone, regardless of how much of an outlier, into a match within around 5 minutes tops. Currently we do still allow up to 20 though.

Part of my preference for 5 minutes vs. 20 or even 10 is that I'd prefer players spend more time in matches than matchmaking. Another reason is that most players, even at the very top, won't wait much longer than 5 minutes and will cancel anyways, so no point in drawing them out that much longer.

We do appreciate your feedback, so keep giving it. Everyone on the team has a huge passion to improve your experiences.

We have found that winrate is ironically not that accurate at predicting whether players will win future matches. They do have some correlation, high-MMR players do tend to have high winrates. But we also have plenty of low-skilled players who played in, e.g., 5qs and have higher win rates than you would expect for their actual skill level.

So we use MMR instead, which is actually very accurate at predicting whether a player will win a match, and play as good as their teammates. Though, keep in mind, every player tilts within about +/-1 ranks from their actual rank during any given match. But, yeah, if MMRs are close, you will probably be happy with them.

Oh, also, sorry, I realized a made an major edit after posting the longer message above, so re-reading it may make some things more clear.

So MMR is equivalent to taking a player's win rate, and then modifying it according to how difficult their matches have been. So in most cases, win% is redundant in the presence of FFA MMR and would just slow MM down. In cases where they disagree, (e.g. 48% win% but high MMR), MMR correctly predicts the winner making the low win% is misleading. Of course if both are low, then the player plays poorly as expected, but that makes it redundant again. Likewise, players with higher win% (e.g. 52%) but low MMR tend to play much worse than their win% would have suggested, which disappoints folks in general.

I think the problem you are running into right now is tightening just how low of an MMR player you can have in your match, even when they are close in Rank. That's something we are investigating.

Fortunately, players in a 5q will not get inflated MMR, that's one of the improvements to the system. So when they go back to solo queue, they will be matched near their MMR still.

We are actively working out improvements to the current premade situation, though I don't have anything to announce at this time.

But, yes, we are very aware of how it currently is working.