yes yes yes yes yes yes and yes
1 Like
One - You get 4k+ players in high plat-low dia games.
Two - Balancing out the match to give each team a 50/50 chance (ASSUMING ALL MMRS ARE ACCURATE) is far better than randomly taking 12 players and putting them on either team.
People will rise and fall to their true rank given enough games regardless of the system used so not sure what your first point was about.
As to the second, you keep saying it’s better, but i don’t see how. With this “better” system" in place, i’d say roughly 50% of the matches i play are one sided. It doesn’t seem like it’s doing it’s job.
I fail to see how this is any different then randomly placing people. Sometimes you get the smurf on your team, sometimes you get the potato. Eventually it all evens out.
The only difference is no one can try to manipulate the system… because you can’t. it’s all w/l based.
5 Likes
There is a very significant difference between balancing the games and not balancing the games that anyone that knows anything about random numbers distribution can tell you and that is variance. For a player that only plays, lets say, 100 games one season the overall balance of games can actually be disastrous. You will actually be very likely to be put on the losing team more than 60% of all your games (the same as someone else will be put on the winning team more than 60% of all their games).
I wrote a little simulation a week or so ago (because I was bored at work and read a comment not very much unlike yours and decided to prove how crazy and bad that idea was) and I’m happy to revisit it and give you some more in depth numbers but just imagine yourself being put in the losing team in like 63 of your 100 matches of the season - would you just tell yourself that “don’t worry if I play 9900 more matches in the end it will sort of even out”?
1 Like
I went on a losing streak a couple weeks ago, lost roughly 17 of 20 games. (over a period of several days so it wasn’t tilt induced.) Don’t tell me everything is balanced now. Like i said, probably 50% of my matches are one sided currently.
You’re telling me that randomly placing people on teams that are the same SR will be worse on average?
Every coin flip has a 50% chance to land on one side. Smurf or potato, it’s random.
To use your example, i could just as easily win 63 of 100 matches. At least we’d know it’s pure random luck, not some shadowy algorithm in the background trying to influence matches.
5 Likes
I’ll just put some numbers out there and then you can draw your own conclusion.
I ran a simulation of a 100 000 made up players all at 2500SR. Each of those 100 000 players played 100 games in a made up season. Each of those games were made up of random players with random SR between 2450 and 2550 - that is some better players and some worse players.
In the system currently being used the teams would get balances by the matchmaker so both teams had the same average SR (well MMR but same thing really) and basically the system would construct each match so that it’s balanced.
However what we instead do for the sake of this simulation is to randomly put these people in two teams so for example you could be in a team with 5 people at 2450 SR and the other team could be 6 people with 2550SR (unlucky I know - might as well just drop and take the penalty and move on to the next match).
Over many thousand of games you would be put on the winning team about 50% of all matches which is good. But here’s what happens in reality:
2 super unlucky people were put on the winning team only 29% of the time (that is on the losing team 71 out of 100 matches)
2610 people were quite unlucky and were put on the winning team 40% or less of the time
17331 fairly unlucky poeple were put on the winning team 45% or less
As you may have guessed already it’s actually quite unlikely to randomly have been put on the winning team exactly 50% of all games - only 8012 people of 100 000 actually were on the losing team exactly as many times as they were on the winning team. A total of 44367 out of the 100 000 made up players in this simulation were actually put on the losing team in more games than they were put on the winning team in this whole made up season.
TLDR; Random teams is bad because personal skill will not matter at all. All that will matter is if you are lucky enough to be put on the winning team more times than you are put on the losing team.
EDIT: I might be trash at this game but I have two masters degrees (computer science and economics) and own a database software company so there are things I actually do know something about - and numbers is one of those things…
3 Likes
By my first point, I was talking about decay, you’d get really high ranked players purposely decaying to smurf really easily, without even needing to lose any games. And MMR does a good job of getting rid of the two things low rank players complain about the most, smurfs and throwers. If it was static W/L smurfs would ruin so many more games because they would take much, much longer to get all the way up to their true rank. Opposite for throwers, they go down much faster, ruining less games because it takes less to get to the bottom. Randomised teams would improve nothing, it has no advantages over the current system. They basically do the same thing, but MMR allows for decay and leaver penalty, and at extreme ranks (bronze / t5) it can balance matches far better.
Agree, I have pretty much always been a good player( mostly talking about console, where I played the longest time on)
I have pretty much always played extremely well in my games( getting the important picks, calling shots, swaping when needed, being positive as hell, getting my job done, as a tank making space protecting my support killing Their support( winston player here) etc… I am pretty much always the one doing clutch plays, making my team cooperate and work as a team, etc…
Im not saying ima god, but , the Mmr, place people that have pretty much the same SR rabking together and balance MMR so teams are fair, so if you are a good player with a good MMR, you will always fall with people that have lower mmr to balance the teams mmr, eich is sad, because it makes you enable to climb like you would deserve, I cant carry the game on my own as a tank like winston, i need good teamate to win, tanks needs a good team, so if you are a high mmr tank, you will most likely fall with lower mmr dps, wich you need to perform well to be effective.
I have done an experiment on one of my account i simply played DPS, getting 3400 in general, on my main acvount playing mostly tank( wich I got more experienced and I am actually better with) I was at 2900 sr, a difference of 500 SR its ridiculous.
I think its one of the reason good tanks have so much problem climbing up the ladder.
So i do think it should be removed.
7 Likes
Maybe you’re just better at DPS?
1 Like
Nahh, I just cant carry as tank, I also play dps on my main account, on dps my placement is quite weird and im not really a god at it.
When im with decent dps( when I duo queu etc… we roll, simply dont have a lot of dps friends, but on stats tracking sites and sites that compares you performance with other players that plays same heroes etc… im generally master or GM( but these sites are not always perfect and precise, but still) compared to dps where I mostly rank diamond)
3 Likes
Couldn’t agree more. I don’t even play on my main account because it is too hard to climb due to the mmr as I’ve had the account since S1.
3 Likes
Yes, this is a great way to simplify the argument!
Can any of the critics answer ForceCmdr’s question? Why do think is MMR beneficial to players?
2 Likes
Because it’s the most accurate way to form fair matches? SR is forced to be relatively consistent in how it changes, especially at Diamond+ where it’s extremely consistent, and such consistently is simply not the most accurate way to measure player skill.
That’s why SR and MMR are separate. SR determines your rewards, so it is subject to rules like “play X games per week” and “change by a consistent amount each win/loss” in order to have a fair reward scheme. MMR determines who goes in what matches, and since the rewards are structured around the separate SR value it can be free to make bolder, but statistically more accurate, changes to the value per game (edit: and by bold, i mean “more than usual” or “less than usual”, not just always more than usual).
2 Likes
Yeah wouldn’t it be great to have GM players and bronze players being in the same game competing for SR? Man it’ll be so fun!
1 Like
Im really interested in a different example. Can you do another run, starting everyone at 2000SR, but with “hidden mmr” distribution the same as what was shown in the recent post by Jeff. Make the game balance matches by mmr but within 100SR (don’t know if you can expand this if no players available for top and bottom). I’m interested to see the final distribution of where the mmr stay sits compared to SR.
Alright folks since Season 8 will be ending in 2 days I won’t have enough time to do all 100 games. I’ve done 50 in 5 days and I’m hoping that will suffice for now.
I can’t tell you how grueling it was playing 10 Comp matches and watching my SR plummet! Here’s my results Cuthbert, Jarlan, and friends, enjoy! LOL
Win or Loss Blue SR Red SR SR After match Points + or –
Day 1
- W 2009 1993 2026 + 16
- W 2022 2007 2047 + 21
- L 2061 2077 2026 - 21
- L 1990 1990 2001 - 25
- W Placement Placement 2024 + 23
- L 2011 2011 2001 - 23
- L 1988 Leaver 2010 1977 - 24
- W 1993 2001 1998 + 21
- W 2005 1998 2027 + 29
- L 2000 2001 2003 - 29
Day 2
- L 1992 1992 1982 - 21
- L 1932 1921 1961 - 21
- L 1953 1955 1939 - 22
- W 1955 1935 Leaver 1959 + 20
- L 1947 1944 1933 - 26
- W 1955 1945 1953 + 20
- L Placement Placement 1933 - 20
- L 1936 1934 1905 - 28
- L 1881 1885 1878 - 27
- L 1885 1893 1853 - 25
Day 3
- L 1837 1848 1830 - 23
- L 1822 1823 1800 - 30
- W 1778 1783 1820 + 20
- W 1815 1823 1844 + 24
- W 1806 1800 1865 + 21
- W 1821 1820 1887 + 22
- W 1826 1824 1907 + 20
- W 1933 1932 1927 + 20
- L 1871 1901 1904 - 23
- W 1894 1892 1926 + 22
Day 4
- L 1860 1825 1904 - 22
- W 1855 1848 1925 + 21
- W 1855 1883 1944 + 19
- W 1947 1952 1968 + 24
- W 1951 1947 1984 + 16
- W 1954 1959 2006 + 22
- L 2001 2013 1980 - 26
- L 1950 1958 1951 - 29
- L 1950 1936 1929 - 22
- L 1997/Leaver 1994 1910 - 19
Day 5
- W 1847 1850 1932 + 22
- L 1907 1902 1908 - 22
- L 1907 1899 1889 - 19
- L 1874 1877 1867 - 22
- L 1833 1819 1841 - 26
- L 1902 1903 1823 - 18
- L 1816 1829 1800 - 23
- W Placement Placement 1824 + 24
- D 1821 1805 1824 0
- L Placement Placement 1798 - 26
4 Likes
My recent post looks awful compared to how it looks in Micrsoft Word and in the Blizzard Forum box right now.
After clicking post it looks worse–I’ll try to clean it up a bit.
There is always option 4, you’re incapable of impartial thinking.
Like I’ve said in other posts, you don’t seem to grasp that the MMR system is totally secret, thus creating an atmosphere of distrust.
You can’t possibly know for sure that Jeff and Co. love you with all their heart and wish nothing but for you to become Grand Master. Perhaps they’re savvy businessmen and have found a way to keep you in your hamster wheel, smiling ignorantly.
3 Likes
Boy I wish I could see your SR right now, puts a lot of forum topics into perspective.
facepalm Jeff doesn’t want me to get to GM because … wait for it … I’m not a GM player.
A ranking system’s purpose is to sort people so that can matched with and against people of the same skill level. It is not to give people a sensation of progress, like leveling up in an RPG.
The one and only way to rank up is to get better as a player and play enough games for the system to measure your improvement.
If you find that you are constantly matched with certain ranks, I hate to break it to you, but that means you are meant to be at that certain rank.
4 Likes