|
On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all.
I don't really understand what you expected, end up with 3100 and complain. I don't know what 500th best venge player on earth(I assume dotabuff rankings) and how it is even possible by playing recent "high bracket" games in your history. Simply you cannot be best 500 with a hero from high bracket at all.
At best you'd be calibrated 3600-3800 band maybe slightly higher but you lost half of the calibration matches. I also assume valve try to be a little pessimistic considering you can abuse FOTM heroes to have a higher win rate so in order to avoid MMR inflation, they may tank a little bit.
All in theory but staying 400-500 MMR away from your "actual MMR" is perfectly normal and I wouldn't count it as a mis calibration at all.
|
9071 Posts
On September 29 2015 19:56 Laserist wrote:Show nested quote +On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all. I don't really understand what you expected, end up with 3100 and complain. I don't know what 500th best venge player on earth(I assume dotabuff rankings) and how it is even possible by playing recent "high bracket" games in your history. Simply you cannot be best 500 with a hero from high bracket at all. At best you'd be calibrated 3600-3800 band maybe slightly higher but you lost half of the calibration matches. I also assume valve try to be a little pessimistic considering you can abuse FOTM heroes to have a higher win rate so in order to avoid MMR inflation, they may tank a little bit. All in theory but staying 400-500 MMR away from your "actual MMR" is perfectly normal and I wouldn't count it as a mis calibration at all.
Its not about miscalibration, the blog was about calibrating as a support. It will probably shock you but the skill of your team mates does matter when you play a support. I complain about being matched with players that made supporting harder and given my poor item choices you can still check in the match history that I was outfarming my cores. It was just an incredibly frustrating experience throughout. And I don't know how I got ranked top whatever with venge I just play a shit tone of games with it and win a lot, I dont think theres another way.
|
I think there is a strong consensus on playing cores are better in terms of game impact and overall MMR gain, maybe except a few(Ler).
I agree mostly and somewhat validated if you look at the top 100 of every region with mostly consist of either core players or pros. I give more credit to 5500 MMR full time support player rather than a 5500 MMR fotm abuser.
I don't believe calibration take into account what hero do you play and try to match it according. They particularly changed the total damage parameter a while ago since people started to abuse zeus ulti all game long.
But it might be that Valve look at a variety of parameters and some of them cannot be achieved as "good" by playing support(I mean simply not prioritize farming and/or playing with farm independent heroes, buy consumables etc..).
Also I strongly think that losing a game, independent of your parameters is a huge negative in calibration matches.
|
On September 29 2015 20:48 Laserist wrote:
I think there is a strong consensus on playing cores are better in terms of game impact and overall MMR gain, maybe except a few(Ler).
I agree mostly and somewhat validated if you look at the top 100 of every region with mostly consist of either core players or pros. I give more credit to 5500 MMR full time support player rather than a 5500 MMR fotm abuser.
I don't believe calibration take into account what hero do you play and try to match it according. They particularly changed the total damage parameter a while ago since people started to abuse zeus ulti all game long.
But it might be that Valve look at a variety of parameters and some of them cannot be achieved as "good" by playing support(I mean simply not prioritize farming and/or playing with farm independent heroes, buy consumables etc..).
Also I strongly think that losing a game, independent of your parameters is a huge negative in calibration matches.
In low ratings I'd say there is no difference between core/support, because although you may not want to admit it, you are all equal skill levels. It's when you get to like 5.7kish when you get games where you are 1k mmr over some people that the core position becomes more important.
|
On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all.
First thing you need to understand: Calibration is not about winning your game. Its about the general impact you have compared to all players in a certain bracket which will then increase your hidden mmr.
An example:
So lets assume you play anti-mage and have 150 cs by minute 30 and 0 hero damage done because you were farming, but 1k tower damage because you were split pushing. Game ends around that mark because your team carried it anyway 4 vs 5.
=> Game won.
=> Game impact = literally zero .
=> Literally zero hidden MMR gained.
=> Next game will be most likely in the same bracket or a bit higher due to you winning.
Lets assume you played the anti-mage game with 300cs by minute 30, 10-0-5 stats, and 15k hero damage and the game ends around that mark but your team threw the game. The MM algorithm will still see that you performed stellar compared to all other players.
=> Game lost.
=> Game impact = Very high.
=> Enormous hidden MMR gained.
=> Next game will be in a much higher bracket compared to the previous one.
|
On September 30 2015 02:24 Ler wrote:Show nested quote +On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all. First thing you need to understand: Calibration is not about winning your game. Its about the general impact you have compared to all players in a certain bracket which will then increase your hidden mmr. An example: So lets assume you play anti-mage and have 150 cs by minute 30 and 0 hero damage done because you were farming, but 1k tower damage because you were split pushing. Game ends around that mark because your team carried it anyway 4 vs 5. => Game won. => Game impact = literally zero . => Literally zero hidden MMR gained. => Next game will be most likely in the same bracket or a bit higher due to you winning. Lets assume you played the anti-mage game with 300cs by minute 30, 10-0-5 stats, and 15k hero damage and the game ends around that mark but your team threw the game. The MM algorithm will still see that you performed stellar compared to all other players. => Game lost. => Game impact = Very high. => Enormous hidden MMR gained. => Next game will be in a much higher bracket compared to the previous one.
Any source or pure subjective experience?
|
Pure subjective experience, which is why sonneiko is 4k mmr player heheh
|
9071 Posts
Well if theres game impact metric for a support my first game of the calibration I went 7-11-19 with 100 LH , 370 GPM, glimmer cape and greaves (I probably healed my team for a shit tone, ww remember) and purchased 15/6 wards and the game was lost. So I'm pretty sure there's no metric that rewards support play during calibration
|
The metric that rewards support play is having a high unranked mmr and winning.
|
On September 30 2015 02:24 Ler wrote:Show nested quote +On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all. First thing you need to understand: Calibration is not about winning your game. Its about the general impact you have compared to all players in a certain bracket which will then increase your hidden mmr. An example: So lets assume you play anti-mage and have 150 cs by minute 30 and 0 hero damage done because you were farming, but 1k tower damage because you were split pushing. Game ends around that mark because your team carried it anyway 4 vs 5. => Game won. => Game impact = literally zero . => Literally zero hidden MMR gained. => Next game will be most likely in the same bracket or a bit higher due to you winning. Lets assume you played the anti-mage game with 300cs by minute 30, 10-0-5 stats, and 15k hero damage and the game ends around that mark but your team threw the game. The MM algorithm will still see that you performed stellar compared to all other players. => Game lost. => Game impact = Very high. => Enormous hidden MMR gained. => Next game will be in a much higher bracket compared to the previous one.
yeah too bad they only determine game impact through things like kda/gpm/xpm/hero + tower dmg etc (all things cores care about)
supports get shafted hard on the 'game impact' scale
|
On September 30 2015 08:48 evilfatsh1t wrote:Show nested quote +On September 30 2015 02:24 Ler wrote:On September 29 2015 19:45 disciple wrote:
So I played a couple of ranked games today with venge. Prior to that I was 500th in the world, just by winning two games I dropped to 97%. I still don't understand why the rank calibration had to essentially reset my progress. I know its shouldn't be hard to get back up in a couple of months but I dont understand the logic behind my calibration match ups. So if its bottom up and you lose the first 2 games of the calibration you are screwed? Cause it never got better, I couldve stopped right after my second game and my MMR would've still calibrated at 3100. I'm just looking for a reasonable explanation thats all. First thing you need to understand: Calibration is not about winning your game. Its about the general impact you have compared to all players in a certain bracket which will then increase your hidden mmr. An example: So lets assume you play anti-mage and have 150 cs by minute 30 and 0 hero damage done because you were farming, but 1k tower damage because you were split pushing. Game ends around that mark because your team carried it anyway 4 vs 5. => Game won. => Game impact = literally zero . => Literally zero hidden MMR gained. => Next game will be most likely in the same bracket or a bit higher due to you winning. Lets assume you played the anti-mage game with 300cs by minute 30, 10-0-5 stats, and 15k hero damage and the game ends around that mark but your team threw the game. The MM algorithm will still see that you performed stellar compared to all other players. => Game lost. => Game impact = Very high. => Enormous hidden MMR gained. => Next game will be in a much higher bracket compared to the previous one. yeah too bad they only determine game impact through things like kda/gpm/xpm/hero + tower dmg etc (all things cores care about) supports get shafted hard on the 'game impact' scale
How do you know this? The fact that they have metrics to judge if your play style is support in the new reborn client leads me to believe it wouldn't be unrealistic for them to use it as d calibration tool as well
|
Well I always heard that account boosters spam zeus or spectre or some other global hero in calibration so that they calibrate at max mmr
|
reborn is very new. i cant speak for mmr calibration now, but for a very long time after ranked was first released, it was pretty evident to everyone that playing support doesnt do shit to get you a good initial rank.
the zeus thing was also abused for a while, although i heard valve changed it up once to stop people abusing hero damage
|
Theres kind of no real way to properly calibrate people for matchmaking, just like MMR is not a strict measure of skill. Not to crap on about it since I already wrote like 7k words on the topic, but MMR is not a measure of skill (ref Day9), its a measure of progress. So how do you properly calibrate a persons instantaneous level of progress at a given time, especially in a game like dota with many variables.
In my opinion if youre calibrating within 1k MMR if where you probably should be then Valve is doing a pretty reasonable job, and if you dont then youre shit out of luck I guess.
|
On September 30 2015 13:01 gaijindash wrote:
Well I always heard that account boosters spam zeus or spectre or some other global hero in calibration so that they calibrate at max mmr
It was changed some time ago. You cannot place 4.5k-5k MMR buy just spam zeus ulti all game long.
|
On September 30 2015 06:35 disciple wrote:
Well if theres game impact metric for a support my first game of the calibration I went 7-11-19 with 100 LH , 370 GPM, glimmer cape and greaves (I probably healed my team for a shit tone, ww remember) and purchased 15/6 wards and the game was lost. So I'm pretty sure there's no metric that rewards support play during calibration
You honestly think 11 deaths is a good performance?
|
On September 30 2015 21:13 schmitty9800 wrote:Show nested quote +On September 30 2015 06:35 disciple wrote:
Well if theres game impact metric for a support my first game of the calibration I went 7-11-19 with 100 LH , 370 GPM, glimmer cape and greaves (I probably healed my team for a shit tone, ww remember) and purchased 15/6 wards and the game was lost. So I'm pretty sure there's no metric that rewards support play during calibration You honestly think 11 deaths is a good performance?
Depends on the average MMR of the match. If it would be a 6k average match with 5 pros, it would be solid.
This is one of the things we are missing while evaluating the MMR algorithm.
|
9071 Posts
On September 30 2015 21:13 schmitty9800 wrote:Show nested quote +On September 30 2015 06:35 disciple wrote:
Well if theres game impact metric for a support my first game of the calibration I went 7-11-19 with 100 LH , 370 GPM, glimmer cape and greaves (I probably healed my team for a shit tone, ww remember) and purchased 15/6 wards and the game was lost. So I'm pretty sure there's no metric that rewards support play during calibration You honestly think 11 deaths is a good performance?
Well I got the least deaths in the team :D
https://www.dotabuff.com/matches/1826192545
|
On September 30 2015 21:13 schmitty9800 wrote:Show nested quote +On September 30 2015 06:35 disciple wrote:
Well if theres game impact metric for a support my first game of the calibration I went 7-11-19 with 100 LH , 370 GPM, glimmer cape and greaves (I probably healed my team for a shit tone, ww remember) and purchased 15/6 wards and the game was lost. So I'm pretty sure there's no metric that rewards support play during calibration You honestly think 11 deaths is a good performance?
death count can really say nothing about a support performance (though it can say a lot about the flow of the game itself)
avoiding deaths isnt a big priority on supports at all (not to say avoiding unnecessary feeding isn't, but there is a such thing as necessary feeding, things like tanking smoke ganks, soaking semi-dangerous exp to not stay underleveled, or initiating when your team doesn't have an initiator who will never die)
the only time supports die very few times is in complete stomps anyhow, one way or the other. you could stomp, or your cores get stomped for being suicidal morons and you just soak exp and survive until you get throned
in the particular game mentioned, it's possible (though unlikely) that a support could have played close to 6-7k mmr level and still gotten 11 deaths in a 55 minute game
|
I agree better players tend to die less but sacrificing yourself to protect a carry shouldn't be a negative thing and this is why evaluating a support from numbers is a tricky business.
|
|
|
|
|
|
EPT
