Here (1). It’s not a perfect test; the sample size is small and there’s no control, but the results show a pretty damning trend. It’s unlikely anyone is going to do a more thorough test though, since that would require ruining a greater number of games to gather the data, and this is already sufficient to indicate there is clearly a problem.
There is simply no alternative explanation for the sudden drop in his games’ average mmr after throwing socials. Certainly his ranked games don’t suggest such a poor performance as to drop him ~400 mmr, so what could be the answer? It’s pretty clearly the socials affecting his ranked mmr. Which, let’s be clear is the value used for MATCHMAKING.
Yeah, no. My experience has been highly consistent on this front, in line with the test cited above.
If I go into socials to do challenges and play like a bot, my next few games will always be massively easier, regardless of playlist. If I go do bots to warm up, games get harder on average, and it takes longer for that to reset back to normal.
Of course, the baseline for this experience isn’t terribly stable because using the less stable mmr to do matchmaking sacrifices a long term balanced experience for short term expected 50/50 “balance,” but it’s still very noticeable.
A few games once in a while randomly is chance. 1-3 games consistently after every streak of poor performance is vanishingly unlikely. It’s so noticeable because as soon as you have a sufficiently good performance the easy games suddenly stop. The mmr likes to adjust up faster than down, so you have to do poorly for longer than you have to do well.
Most below average players aren’t generally that invested in the nuts and bolts of the game they aren’t good at, by the simple nature of the relationship between time investment and skill level. Sure, there’s some deviation on that trend, but it will hold true on average.
Consider also that showcasing this would be a miserable experience. Making a brand new account for instant onyx games is one thing (2), but placing at your true rank and then subsequently boosting your mmr so you can climb is an entirely different beast.
You would have to go slay bots for x games, then go into ranked to get stomped and pray you get someone on your team who can carry the 3v4 so you can get the boosted csr gains for that game. Which is unlikely, because your higher mmr will guarantee your teammates are below the mmr average for the lobby, since you are the expected carry to make the game a 50/50. So you endure this miserable grind all so that when you finally get that onyx badge you can have the awful experience of slowly dropping back to gold.
The alternative is to go feed as hard as possible in swats or something for 10+ games, stomp lobbies below your mmr, and get whatever reduced gains which I assume will diminish to 0 at some point because your mmr will need to be so far below onyx to get those easy games that the system just won’t give you anything.
Not a lot of people lining up to test that.
It’s not harmless, because some people want to use the bots to warm up, and it affects their ranked experience. Or a new player might do a few bot games and immediately go to ranked to get this experience (2), and immediately stop playing. Edge case though the example may be, the underlying problem exists regardless of the severity, and has an impact across the board. Even if most individuals don’t notice it, it degrades the experience by some margin by reducing the overall fairness of the matchmaking.
Doesn’t look like it (3). Seems to me that his mmr was barely dropping, if at all, for all 12 games. His team is consistently around 15-1600 going up against 19-2000+, and half the time his team has a higher average mmr. Then the reset happened and his lobbies, like everyone else’s, are all over the place and it’s hard to draw any further conclusions.
I see no evidence whatsoever that his mmr was rapidly plummeting, it looks more like it was still glued to the roof from all those bot games until the reset hit and did whatever they did.
For it to be the matchmaking factor so that we don’t have this miserably discouraging experience.
If I belong at a higher rank, don’t force my games to a 50/50 winrate immediately and slowly edge my csr upwards as I grind my way through games that are stacked against me to balance out the mmr. Just match me at my display rank and let me climb (or fall) to the level where my display rank matches my level of play, using the mmr as a weight to bring my rank in line with my skill level faster.
Yes, because having a display rank that doesn’t match the lobbies you get placed in is disheartening and makes the game less fun to play. Obviously the csr number will never be an exact representation of player skill, but the point of that number is to showcase a player’s approximate position in the skill curve. If nothing else, it tells you where you are at.
So to have that number be radically distant from the average of your lobbies in is pretty disheartening. It’s not fun having your display rank say you are the worst in the lobby because your mmr is higher than your csr, and matchmaking for immediate 50/50 with the mmr means you are also forced to try to carry games, despite being the lowest rank in the lobby.
If the csr number displays something less than my actual skill, let me prove it by matching me against others at the same csr. Then use my mmr, as the game’s interpretation of my true skill, to weight my csr adjustments to keep that display number as close to accurate as is reasonably possible.
That way, not only will match balance be less chaotic by using a more stable number for balancing purposes, but rank can serve as an indicator for relative performance over time, allowing players to track their improvement. Which is kind of the entire point.
No. Your mmr is a number that can be plotted as a probability density against all other mmr values in a population to give a gaussian distribution. You don’t have to use the gaussian, for example chess used to use gaussian distribution but now they use the logistic distribution, because it’s a better predictor of performance. Regardless of what probability density function you use, mmr is a number used as a numerical approximation of skill. The same is true of csr. The only difference between them is that mmr can move more freely than csr, which is tied to the win/loss parameter as well as to mmr.
The point of having a hidden mmr is generally to track the moving average of player performance and use that as a weight to modify rank gain/loss. The point of doing that is to create a buffer against the random variables that can affect win/loss in games that aren’t a pure 1v1 with (mostly) unchanging start conditions. This is why games like chess don’t need mmr and can get away with just the one number. And to be fair, you can do the same for multiplayer games, which some mobas do, you just need a more complex elo calculation to account for the more complex conditions.
- https://www.reddit.com/r/halo/comments/sjcg8u/its_official_matchmaking_is_broken_social_games/?utm_source=share&utm_medium=web2x&context=3
- https://twitter.com/Gamesager/status/1484630839707111427
- https://leafapp.co/player/Fabled+Fella/matches
