Concede it is a separate MM, they just did not ignore previous data that was collected. This gave people a proper starting point whether or not they stay there is determined on how accurate the prediction was. There is no reason to wipe the slate clean because I for one do not want to start in the dota trash league and I dont think any other decent player does as well.
Ya. i agree.
Your ranked MMR is pretty much determined prior to your 10 calibration games. can get 4000+ even with 10 loss and some people get 2000+ with 10 win.
6 wins - 4 losses. Started at about 3830 mmr, got up to 3890 till game 9 and stomped hard with BH the last one and it skyrocketed to 3960. Is this actually good at all?
Went 7-3 during calibration for Solo play. Games I lost were more the result of the generic "Chooses Sniper and takes 40 minutes just to get the Broadsword for his Shadowblade" style team mates. Played an absolute shit match as QoP but I'm certain we were going to lose it anyway since my entire team went random and they counterpicked us.
Resulted in a 3502 MMR. Apparently this means I'm not in the shitlord 2k hellzone but not exactly top tier player that will dominate nearly every match with any hero. Which is about correct in my evaluation of my own skills.
They said that they anticipate Ranked to be higher, but they don't know what yet. At any rate, that puts you at approximately the top 5% of pub players, so yeah, congrats. For ranked I'd assume you're probably around top 15-20%.
Dendi is rated at 5600 solo queue so thats a clear indicator that the old metric is slanted by something like 1500 more points than the original metrics.
I don't get why they don't make mmr data more accessible. Yes alright, there will be always exploiter but for the sake of the majority, would be good to get more information
With the removal of the dota_game_account debug information, it seems like the MMR analysis will have to come to an abrupt ending.
Based on my limited experience with Team MMR, an individual's rating will converge towards the stack's average rating, regardless of the outcome or individual contribution towards the game. This means your MMR will increase if you are on the low end of the stack and decrease if you are on the high end. Outcome and/or contribution can accelerate this convergence, but there doesn't seem to be a way to avoid it.
If your goal is to climb the MMR ladder, then a good rule of thumb to follow is to stack with players rated higher than yourself (giving you a chance to increase in MMR), and never stack with lower rated players than yourself (virtually guaranteeing MMR loss).
Below are the data points I collected, using the format: Game #, Result, KDA, Rank, Delta from last rating, Uncertainty. Not listed are the ratings of the people I stacked with or a generalization of how the game went. Most of the groups were either 3-stacked(3100, 4000, 4100) or 2-stacked (3500, 4000). I picked Earth Spirit for each of the games.
My 7 first ranked games (calibration) was fun, not unbalanced. But since Monday, i get the worst games since i play Dota. I really never saw that.
Is the fact that the games take 5 min to start with a extended search criteria? No idea, but for moment he seems better to wait for playing in ranked match, time that everything is secure..
KDA doesn't seem to make any difference in a game you lose, even when you play well and lose it narrowly. Went 8-4-31 with a support and lost the game. Had best KDA among the 10 players in the game, but the MMR dropped 25 points.
In a game where I might play terribly, have a much worse KDA but win, I am sure my MMR will go up. So the Ranking system appears far too heavily weighted towards overall win rate.
Thanks to my unfairly low rating, I'm getting paired with players that play like faeces flinging monkeys randomly bashing keyboards, so the best way for me to improve my MMR is to abandon my usual support role and play a solo mid hero to try to stomp the game alone.
Okay, Ive played like 30 games ranked now, and Ive seen no indication that anything other than win / loss matters. Every single game its either +25 for win or -25 for loss, regardless of performance. Ive had 17-0-16 and 5-2-36 games where I only earned the usual 25, and what Id say is almost flawless support games on losing teams where I lose 25 rating.
Taken this into acount, that there is no individual skill factor, its insanely slow to gain rating if solo queuing. A player who manages a consistent 52% winrate (fairly good) solo will only gain 100 rating every 100 games (52 wins and 48 losses = net gain of 4x 25 =100) . To get up from 3800 to 4300 takes 500 games with this system. Very disappointing and the only solution is to pick ursa wisp or similar super stompy combos to boost rating until you get away from shit tier.
Im by no means a great player but I have friends of similar skill who are 4400-ish while Im 3800 (maybe because Ive stacked a lot with shit tier friends before ranked came out and the system thinks I am one of them, I dont know) and the difference in game quality when I stack with those friends compared to playing alone is staggering. Knowing that I will have to play ~500 games to get my rating up to a decent level solo is a huge downer..
"Im by no means a great player but I have friends of similar skill who are 4400-ish while Im 3800 (maybe because Ive stacked a lot with shit tier friends before ranked came out and the system thinks I am one of them, I dont know) and the difference in game quality when I stack with those friends compared to playing alone is staggering. Knowing that I will have to play ~500 games to get my rating up to a decent level solo is a huge downer.."
I played ALOT (and still do) with players who are just being introduced to the game or play extremly seldomly.
In those games i keep playing support as i usually do in ranked, that means i don't get to carry the game to victory that easily as i would with a mid hero carry etc.
that impacted quite a lot on my initial rating, while as you said i am also by no means a great player i still believe that the rating system doesn't allow you to progress properly, especially if you are playing a support role.
We can hope that in the future the system gets a bit perfected by what i would also really like is for Valve to get out of the standard routine and cooperate more with the community by showing what the rating takes precisely into account, knowing ofc that we will ALWAYS have exploiters but its a shame for that to not allow the good guys to understand the system better.
Well, thats pretty normal isn't it? When you have played 30 games and your MMR is 3800, even if you perform at 5700 MMR on your 31st game you are only going to increase to (30(3800) + 5700) / 31) which is 3861. If you really want to get to another tier, you should improve first then create another account. Current your standard is only at the 3800 bracket.
My standard is not at the 3800 bracket though. I know most here will probably see my post and think "lol 3800 shitter who thinks hes good, Dunning-Kruger effect lol l2p" but I have a good grasp of the game and perfectly honest in my solo queue games I am the one on my team with best game sense more than 90% of the time.
I mean, just yesterday i got matched with Krobe who went mid with Poor Man's Shield first item, and then built Vanguard as second item. Support players in my games constantly feed because they are out of position and then say "help me ffs" when they die to nyx for the 10th time warding on their own instead of waiting for a good opportunity like Nyx being dead, and if I support, the carries get their space but then make horrible decisions mid and late game and throw the game, despite having ward vision and invis detection 100% of the time.
Im not saying I should be 5k +, I shouldnt. But I should be at least 4200 judging from the ratings of friends that are on equal skill level as me. And that will take me 400 games with the current system, even if I am the best player on my team every single game.
Most dota players think that they are better than they actually are. I have gained rating after losing once. Even players at 4.5k rating have to deal with shitty team mates. When you solo queue, you should know how to take calculated risks, and when not to take risks depending on the situation. Players at higher MMR know what to do when they are teamed with shitty players and that's something that a good player should learn in order to play better in solo queues.
I dont disagree with any of that, what I do disagree with is Valve's system for rating gain / loss, which makes it take several hundred games to gain rating.
When something clicks for a player and they actually become better, it takes a stupid amount of time before the rating reflects it.
Well, i'm sure the increase of your MMR will increase more when you start playing much better consistently. Before MMR was implemented, i played on another account which was stuck at the Normal bracket with 600 games played. I wanted to play in that account for the fun of owning players who are much lesser skilled. I managed to play the account's first Very High game after approximately 20 games. I guess valve's MMR algorithm also will factor in the fact that players who improve tremendously overnight will have a higher gain in rating as they show that they play at a much higher skill level consistently.
I find the quality of a game is quite proportional to the amount of players who have a dotabuff profile. Generally you need at least 4 for a decent game or else you just end up with one guy feeding like mad and it's just gg. I'm not saying the players are higher skill, just that they are less likely to be of awful skill / feed and be annoying.
I think that is something people have against the ratings. A guy feeds 5 kills in 5 minutes to the enemy carry and you lose rating because you "didn't do enough". There is really very little you can do in such a situation.
Also wanted to ask, how do you guys feel about MMR going up when you lose games (apparenly supposed to happen if you play well) I was carrying my team in one game, srsly doing everything doing everything intiating etc KDA 11.5 and my MMR still dropped after the game. So does this really happen? (MMR can go up after lost games that is)
I've never seen it go up after a loss. I mean I usually don't have great kda after a loss, but I had a 7/4/10 naix game with decent farm for a losing side and still dropped the same as my terrible games. Seems to go up and down about 20 every win/loss.
I honestly don't give a crap what my mmr is, I just don't want to play with fucking awful people anymore. I get people who don't communicate, choose the dumbest builds, feed like crazy etc etc. I'm not that good but I don't do stupid shit. I might just go back to normal matches, I wasn't having this problem there.
I doubt MMR increases in a loss game. The best I played in a lose game decreased by around 3-5.
Anyway, my solo queue rating got 4011 with 3W/7L, party rating 4045 with 6W/4L. So far I haven't been matchmade with really noob players, everyone played very equally. Love this new feature :)
It's hard to gain points when u lose a game, but think of it.
If 10 players with an average MMR of 5000 are playing against each other, and the one with the lowest MMR on the losing team tends to go mad, ie, play gd, and still lose, then he might actually gain some points, but does that happen?
I doubt it.
yes i jsut had a solo game with a 4 stack in my team and they got so mad at me at the end...hahaha of course blame the solo alch who went mid not the 134 last hits in 35 min magina :DD
eternal envy is retard. You DO need boots on weaver unless you have your whole team backing you up, and even then you should have early boots
3415, 7/10 wins mostly death prophet
^
4W/6L actually, counted the party ranked games last time
Format: Game #, Result, KDA, Rank, Delta from last rating, Uncertainty
Team MMR:
1, Win, 6-3-12, (?), (?), (?)
2, Win, 6-3-24, 4042, (?), 689
3, Win, 9-3-14, 4037, (-5), 660
4, Win, 4-4-8, 3998, (-39), 651
5, Win, 8-4-13, 3997, (-1), 616
6, Loss, 10-8-20, 3923, (-74), 609
Solo MMR:
1, Loss, 6-10-20, 4036, (?), 668
2, Win, 10-3-17, 4089, (+53), 646
do we win more points if we win 4v5 ?
Solo mm doesn`t make any sense, I have 4400 mmr solo and played vs 2 guy 5200 and 1 guy 3600.
2774 solo rating, won 9 out of 13 games played and has barely moved since the first game after which it was 2771.
Placement: 5 wins & 5 losses - 4971 solo MMR
Most games were pretty balanced and really enjoyable.
5490 is ok?
@ sampson
because ranked mmr was suppose to be separate from normal MM
This way, they could have just made a new "unranked" mode, and shown us what our normal MMR rating was all along.
3 men stack . 4 win - 6 loss. got 3600 party mmr. (other two frd got 3300 and 3000)
So I guess it takes kda into account.
then i got another 1 win and that number doesn't change. (I did kda = 1/3/0) . ya. kda matters in mmr.
Concede it is a separate MM, they just did not ignore previous data that was collected. This gave people a proper starting point whether or not they stay there is determined on how accurate the prediction was. There is no reason to wipe the slate clean because I for one do not want to start in the dota trash league and I dont think any other decent player does as well.
Format: Game #, Result, KDA, Rank, Delta from last rating, Uncertainty
Team MMR:
1, Win, 6-3-12, (?), (?), (?)
2, Win, 6-3-24, 4042, (?), 689
3, Win, 9-3-14, 4037, (-5), 660
4, Win, 4-4-8, 3998, (-39), 651
5, Win, 8-4-13, 3997, (-1), 616
6, Loss, 10-8-20, 3923, (-74), 609
7, Loss, 11-13-19, 3878, (-45), 572
Solo MMR:
1, Loss, 6-10-20, 4036, (?), 668
2, Win, 10-3-17, 4089, (+53), 646
Ya. i agree.
Your ranked MMR is pretty much determined prior to your 10 calibration games. can get 4000+ even with 10 loss and some people get 2000+ with 10 win.
6 wins - 4 losses. Started at about 3830 mmr, got up to 3890 till game 9 and stomped hard with BH the last one and it skyrocketed to 3960. Is this actually good at all?
Went 7-3 during calibration for Solo play. Games I lost were more the result of the generic "Chooses Sniper and takes 40 minutes just to get the Broadsword for his Shadowblade" style team mates. Played an absolute shit match as QoP but I'm certain we were going to lose it anyway since my entire team went random and they counterpicked us.
Resulted in a 3502 MMR. Apparently this means I'm not in the shitlord 2k hellzone but not exactly top tier player that will dominate nearly every match with any hero. Which is about correct in my evaluation of my own skills.
what's the console commend to see mmr ?
Developer 1
Dota_game_account_debug
BogiDotO, you will have high uncertainty in the first calibration matches. That means, you get matched with a wider range of players.
DEVi[A]NCE
These are the pub scores as released by Valve.
5% 1100
10% 1500
25% 2000
50% 2250
75% 2731
90% 3200
95% 3900
99% 4100
They said that they anticipate Ranked to be higher, but they don't know what yet. At any rate, that puts you at approximately the top 5% of pub players, so yeah, congrats. For ranked I'd assume you're probably around top 15-20%.
Dendi is rated at 5600 solo queue so thats a clear indicator that the old metric is slanted by something like 1500 more points than the original metrics.
Dendi is 5.7K btw
i dont get this ...
i played with a friend in group.
10 matches, 7win, 3 looses.
2331,
now is this good or bad ?! i read alot of solo, pub, etc ...
That's average, but if you compare with most people here, it's pretty low.
did they take away the ability to see uncertainty?
yes u cant see your rank too until 10games completed
:/ that sucks. I had an excel file where I charted my solo/team rank/uncertainty.
4627. was 6-1, ended 6-4. A little drunk during the last 3 games and fed like hell lol
they seem to have removed the display of ranking with dota_game_account_debug ... lame.
I don't get why they don't make mmr data more accessible. Yes alright, there will be always exploiter but for the sake of the majority, would be good to get more information
With the removal of the dota_game_account debug information, it seems like the MMR analysis will have to come to an abrupt ending.
Based on my limited experience with Team MMR, an individual's rating will converge towards the stack's average rating, regardless of the outcome or individual contribution towards the game. This means your MMR will increase if you are on the low end of the stack and decrease if you are on the high end. Outcome and/or contribution can accelerate this convergence, but there doesn't seem to be a way to avoid it.
If your goal is to climb the MMR ladder, then a good rule of thumb to follow is to stack with players rated higher than yourself (giving you a chance to increase in MMR), and never stack with lower rated players than yourself (virtually guaranteeing MMR loss).
Below are the data points I collected, using the format: Game #, Result, KDA, Rank, Delta from last rating, Uncertainty. Not listed are the ratings of the people I stacked with or a generalization of how the game went. Most of the groups were either 3-stacked(3100, 4000, 4100) or 2-stacked (3500, 4000). I picked Earth Spirit for each of the games.
Team MMR:
1, Win, 6-3-12, (?), (?), (?)
2, Win, 6-3-24, 4042, (?), 689
3, Win, 9-3-14, 4037, (-5), 660
4, Win, 4-4-8, 3998, (-39), 651
5, Win, 8-4-13, 3997, (-1), 616
6, Loss, 10-8-20, 3923, (-74), 609
7, Loss, 11-13-19, 3878, (-45), 572
Solo MMR:
1, Loss, 6-10-20, 4036, (?), 668
2, Win, 10-3-17, 4089, (+53), 646
3, Win, 7-3-11, --no data available due to removal of debug information--
My 7 first ranked games (calibration) was fun, not unbalanced. But since Monday, i get the worst games since i play Dota. I really never saw that.
Is the fact that the games take 5 min to start with a extended search criteria? No idea, but for moment he seems better to wait for playing in ranked match, time that everything is secure..
I had this too yesterday afternoon. I hope the serious players were at work or I got unlucky twice in a row or something :D
KDA doesn't seem to make any difference in a game you lose, even when you play well and lose it narrowly. Went 8-4-31 with a support and lost the game. Had best KDA among the 10 players in the game, but the MMR dropped 25 points.
In a game where I might play terribly, have a much worse KDA but win, I am sure my MMR will go up. So the Ranking system appears far too heavily weighted towards overall win rate.
Thanks to my unfairly low rating, I'm getting paired with players that play like faeces flinging monkeys randomly bashing keyboards, so the best way for me to improve my MMR is to abandon my usual support role and play a solo mid hero to try to stomp the game alone.
Same here, tuesday and wednesday matches were horrible, while initial matches were tough and well balanced.
the last two ranked games I had were very enjoyable and the team was really cooperating, especially on the last one.
Okay, Ive played like 30 games ranked now, and Ive seen no indication that anything other than win / loss matters. Every single game its either +25 for win or -25 for loss, regardless of performance. Ive had 17-0-16 and 5-2-36 games where I only earned the usual 25, and what Id say is almost flawless support games on losing teams where I lose 25 rating.
Taken this into acount, that there is no individual skill factor, its insanely slow to gain rating if solo queuing. A player who manages a consistent 52% winrate (fairly good) solo will only gain 100 rating every 100 games (52 wins and 48 losses = net gain of 4x 25 =100) . To get up from 3800 to 4300 takes 500 games with this system. Very disappointing and the only solution is to pick ursa wisp or similar super stompy combos to boost rating until you get away from shit tier.
Im by no means a great player but I have friends of similar skill who are 4400-ish while Im 3800 (maybe because Ive stacked a lot with shit tier friends before ranked came out and the system thinks I am one of them, I dont know) and the difference in game quality when I stack with those friends compared to playing alone is staggering. Knowing that I will have to play ~500 games to get my rating up to a decent level solo is a huge downer..
"Im by no means a great player but I have friends of similar skill who are 4400-ish while Im 3800 (maybe because Ive stacked a lot with shit tier friends before ranked came out and the system thinks I am one of them, I dont know) and the difference in game quality when I stack with those friends compared to playing alone is staggering. Knowing that I will have to play ~500 games to get my rating up to a decent level solo is a huge downer.."
I played ALOT (and still do) with players who are just being introduced to the game or play extremly seldomly.
In those games i keep playing support as i usually do in ranked, that means i don't get to carry the game to victory that easily as i would with a mid hero carry etc.
that impacted quite a lot on my initial rating, while as you said i am also by no means a great player i still believe that the rating system doesn't allow you to progress properly, especially if you are playing a support role.
We can hope that in the future the system gets a bit perfected by what i would also really like is for Valve to get out of the standard routine and cooperate more with the community by showing what the rating takes precisely into account, knowing ofc that we will ALWAYS have exploiters but its a shame for that to not allow the good guys to understand the system better.
Well, thats pretty normal isn't it? When you have played 30 games and your MMR is 3800, even if you perform at 5700 MMR on your 31st game you are only going to increase to (30(3800) + 5700) / 31) which is 3861. If you really want to get to another tier, you should improve first then create another account. Current your standard is only at the 3800 bracket.
My standard is not at the 3800 bracket though. I know most here will probably see my post and think "lol 3800 shitter who thinks hes good, Dunning-Kruger effect lol l2p" but I have a good grasp of the game and perfectly honest in my solo queue games I am the one on my team with best game sense more than 90% of the time.
I mean, just yesterday i got matched with Krobe who went mid with Poor Man's Shield first item, and then built Vanguard as second item. Support players in my games constantly feed because they are out of position and then say "help me ffs" when they die to nyx for the 10th time warding on their own instead of waiting for a good opportunity like Nyx being dead, and if I support, the carries get their space but then make horrible decisions mid and late game and throw the game, despite having ward vision and invis detection 100% of the time.
Im not saying I should be 5k +, I shouldnt. But I should be at least 4200 judging from the ratings of friends that are on equal skill level as me. And that will take me 400 games with the current system, even if I am the best player on my team every single game.
Most dota players think that they are better than they actually are. I have gained rating after losing once. Even players at 4.5k rating have to deal with shitty team mates. When you solo queue, you should know how to take calculated risks, and when not to take risks depending on the situation. Players at higher MMR know what to do when they are teamed with shitty players and that's something that a good player should learn in order to play better in solo queues.
I dont disagree with any of that, what I do disagree with is Valve's system for rating gain / loss, which makes it take several hundred games to gain rating.
When something clicks for a player and they actually become better, it takes a stupid amount of time before the rating reflects it.
Well, i'm sure the increase of your MMR will increase more when you start playing much better consistently. Before MMR was implemented, i played on another account which was stuck at the Normal bracket with 600 games played. I wanted to play in that account for the fun of owning players who are much lesser skilled. I managed to play the account's first Very High game after approximately 20 games. I guess valve's MMR algorithm also will factor in the fact that players who improve tremendously overnight will have a higher gain in rating as they show that they play at a much higher skill level consistently.
I find the quality of a game is quite proportional to the amount of players who have a dotabuff profile. Generally you need at least 4 for a decent game or else you just end up with one guy feeding like mad and it's just gg. I'm not saying the players are higher skill, just that they are less likely to be of awful skill / feed and be annoying.
I think that is something people have against the ratings. A guy feeds 5 kills in 5 minutes to the enemy carry and you lose rating because you "didn't do enough". There is really very little you can do in such a situation.
Got 7-3 MMR about 3500 feelsbadman.jpg Watched Dendi he has 5864 :( oh well noobness revealed
Also wanted to ask, how do you guys feel about MMR going up when you lose games (apparenly supposed to happen if you play well) I was carrying my team in one game, srsly doing everything doing everything intiating etc KDA 11.5 and my MMR still dropped after the game. So does this really happen? (MMR can go up after lost games that is)
I've never seen it go up after a loss. I mean I usually don't have great kda after a loss, but I had a 7/4/10 naix game with decent farm for a losing side and still dropped the same as my terrible games. Seems to go up and down about 20 every win/loss.
I honestly don't give a crap what my mmr is, I just don't want to play with fucking awful people anymore. I get people who don't communicate, choose the dumbest builds, feed like crazy etc etc. I'm not that good but I don't do stupid shit. I might just go back to normal matches, I wasn't having this problem there.
I doubt MMR increases in a loss game. The best I played in a lose game decreased by around 3-5.
Anyway, my solo queue rating got 4011 with 3W/7L, party rating 4045 with 6W/4L. So far I haven't been matchmade with really noob players, everyone played very equally. Love this new feature :)
It's hard to gain points when u lose a game, but think of it.
If 10 players with an average MMR of 5000 are playing against each other, and the one with the lowest MMR on the losing team tends to go mad, ie, play gd, and still lose, then he might actually gain some points, but does that happen?
I doubt it.
People gets x10 mad when they lose ranked matches so fuck this shit i am playing normal games again.
People who seems nice at the begining of the game gets crazy when we lose.
yes i jsut had a solo game with a 4 stack in my team and they got so mad at me at the end...hahaha of course blame the solo alch who went mid not the 134 last hits in 35 min magina :DD