This could be a “blame the MM system” thread, but there’s plenty of them already. I just wanted to bring this interesting fact.
And, ya, matching 5 stack masters MMR against 2 stack low silvers is pretty problematic. I wonder how boring it would be to win around 98% of my games.
These aren’t my games. They’re from the highest MMR player in QM right now. He’s at 3300.
As for me, I’m content with my 3100. For exclusively solo queuing, it’s good.
As for him, he doesn’t seem to have a difficult time to find games, by looking at his match history.
Accourding to heroesprofile my NA QM is of masters, I played 1 qm today , these were gold max
Also my personal ranked stats on both EU and NA are lower then what they really are
Ranked is the only accurate stat I can tell what I am
Accourding to heroesprofile we also had the self claimed #1 wm and medivh in the world here in the forums
Given that you’re using an external website to prove it, yes, it sounds crazy.
er, is there a problem?
I don’t see a problem with a third party website that doesn’t capture even close to half of all games played not having an accurate MMR rating for players or calculation for MMR indicating any probably for HotS.
Is the interesting fact meant to be that websites that are third party are unable to accurately measure your MMR?
Because while factual, that’s been well known since, well, forever?
yea, not being able to see QM MMR leads to some head-scratchers.
I honestly still don’t get why they don’t open things up to all MMR levels.
I’ve played with/against players from Master/GM to bronze 5 in QM based on ingame rankings.
It is safe to say that QM still could use a bit of clarity.
Ah yes, I remember that guy.
Doomguy.
With 452 games played in 2020 season 1, but with only a hero MMR of 2682 while the top in QM was 3298.
Unfortunate.
I think that personal MMR is more interesting to know than having a certain rank. You know why? The #1 EU with 37 wins this season has only 3051 MMR while the #2 (Cris, the absolute God) has 3379 MMR with 215 wins.
3rd party sites aren’t the matchmaker; they didn’t make the match, they only know that it was made and try to adjust accordingly
Since uploading is voluntary, different accounts have different levels of confidence with the system; when stuff happens outside it’s expectations (due to divergent confidence) it can only retroactively make changes as it gets more reactive information on all players involved
QM has certain quirks that 3rd parties sites don’t know, so they don’t account for those accordingly
Having more games uploaded on a site tends to inflate the stats compared to those shallow representations; this can further skew data in modes that don’t display relative mmr (ranks) and have people guess entirely on what the system is doing.
What you claim as a ‘fact’ really isn’t; it’s mostly visual conjecture slapped through an implied reference on links that may not age so well as more games are added to any of the players on those losing sides.
On one hand, it’s just Heroes Profile estimation.
On the other, I suppose it’s either a funny edge case or the site “thinking” it’s impossible to have this matchup unless in-game your MMR is lower.
it would depend how the calculation is done.
It is possible to make a system that does that, but especially in SL, where the matchmaking rules have a hard limit of how they’ll expand the matchmaker, i would say it is improbable that it could happen.
In QM or URD with super long queues, I suspect it might be possible. But that depends how exactly their MMR system works.
Even in those super long queues, you would than need a terrible mismatch where one team is much better than the other.
There are basically two options.
If you always have a 51% win chance, then over 100 matches your score shall be 51-49 and you shall gain no points. To achieve that, 400 points have to be deducted, meaning 4 per match. HP uses different numbers but the concept is the same I suppose.
You can express it as a multiplication. So your earnings are
200 * (Win ? 1 : -1) + (0.5 - WinChance) * 400
You can extract 200 as a constant.
But how do you figure out that 51%?
It’s derived anyway, so you can mimic the same based on rank difference, which is slightly easier if you also want to involve team compositions and somesuch. You could simply subtract the MMR difference and while at it, add team size adjustments, too (HOTS does). With this you can express a 101% (or -5%) chance to win, which comes out of the function above.
It wouldn’t be a first. Economists often replace multiplication with addition (1.05 * 1.05 => 1 + 0.05 + 0.05 = 1.1), but of course as the above example shows, such simplifications only works between sensible boundaries. We all know 1.5 * 1.5 != 2. (2.25) I only saw it in American economic books so I wouldn’t be surprised if it would be standard US education to simplify like this. Blizzard is American, so.
It’s not impossible and it depends on the system being used. While blizzard’s mmr is based on the elo system, we don’t know the exact details and there are variations of elo used in other games that are known, so they could use an amalgamation of those.
Since ranked points adjust after each game, we assume that the w/l is evaluated per game but depending on how “confidence” works, HotS may use something closer to tournament ELO or even WHR (whole history rating) In some events for ELO rating, the assessment of rank between players isn’t based on each win/lose, but on the expected score a player got after the event compared to what they actually received. If their ranking in relation to those they played against suggests that have x% probably to win, then that’s how many games in a series they are expected to win based on what they faced. In the case of chess, that evaluation accounts for Draws, while draws are generally improbable in HotS, they are possible and I don’t remember how the game treats it, but the system may have to account for it.
Digression aside, if the mmr estimations aren’t strictly win/lose after each game, but after a specific series of games (the designated ‘event’) then the net result after 1 specific game may end up as an overall loss of mmr despite the last game in the set being won.
That is to say, maybe the system evaluates harder on sets of 10 games: say a player won 5 games out of the ten, but the system expected 6 wins, so despite the 10th game being a win, it was still shy of the 6th, so the internal change is downward at that specific point. Similarly, since some ranked seasons were defined by unwanted jumps in ranking, bugs and other assessment variables can always be a factor, they just may not be intended.
I think hots used to have WHR (whole history rating) in 2015-2017, when you could gain as little as 0 points per win (or even a few negative ones) if there was an abnormally high MMR discrepancy between teams and your MMR certainty for the season was still low (not enough games played).
After SL was introduced, you seem to get 150 points even for winning someone multiple leagues below you in average MMR which just seems wrong (enemy favored + 50)
I remember getting my butt kicked by high Rank 1’s (much different from average rank 1’s) in preseason, and our team only lost 0 pts, not 150. That felt much more fair.
a note, remember that until SL (or was it in the last TL season?) hit, the rank didn’t reflect your MMR. so while your rank may not have moved, it is unlikely that your actual MMR was negative in either of these scenarios.
I remember that. Like, you had a Plat 1 guy. He was banning heroes, even if your team had a bunch of Diamond 4’s and 3’s. Just because his MMR was higher than his rank and it hadn’t caught up.
Happier with the current system, where highest points is automatically the highest MMR and bans stuff.
I had this argument the otherday and I was called an elitest cause i wanted to play with and against players of equal skill level instead of making a smurf and going to bronze and destroying everyone. Like that is a huge problem and its why the matching in this game is so bad.
