|
28078 Posts
On November 01 2014 17:52 Unleashing wrote: Top 1% if that thing valve posted way back can be trusted. That was unranked MMR. Although at the start ranked would have been somewhat close. But it's changed since then with inflation/more players/ people getting better.
|
I don't quite believe there is an inflation you have to remember in games with people that abandon there's a reduction in overall MMR in the pool. Because only 4 people gain MMR while 6 people lose it.
|
Why inflation?
5 people always gain MMR, while the other 5 lose it. Always cancels out. Unless you're playing against players who haven't got calibrated yet.
Not to mention the abandons, where less people gain MMR over the ones who lose MMR.
|
On November 02 2014 01:38 trinxified wrote: Why inflation?
5 people always gain MMR, while the other 5 lose it. Always cancels out. Unless you're playing against players who haven't got calibrated yet.
Not to mention the abandons, where less people gain MMR over the ones who lose MMR.
Smurf accounts, new accounts.
They have the level requirement to limit the mmr impact from people being incorrectly ranked when they enter ranked. Remember the patches where you could calibrate at 6k and people sold those accounts? MMR inflation since you gained that rank in unranked and then lost 3k mmr in the next 300 games. Thus you just added 3k to the total pool before stabilising.
|
Yeah but that doesn't factor in the many people that place below the average MMR or the MMR lost with abandons. I don't believe there is an inflation, not that any of us can prove it Only valve would know that.
|
People getting better, more people playing, etc. don't lead to inflation. That just means that if 4300 is 1%, that 1% will be objectivelly better or a bigger amount of people. Games often have stuff that leads to inflation, but it's hard to say if, by what and by how much the MMR in Dota is really affected by it.
|
Hey guys, not sure where I should post this, but I was wondering if anyone could post a few tips on maintaining the creep equilibrium as a carry, esp. if I have a shit support who keeps pushing the lane out. Any answers would be much appreciated!
|
If your support keeps pushing the lane you're kinda fucked unless you can get someone to pull. If not you just gotta deny when ever possible and only hit creeps for the last hit.
|
What Jasper said, sometimes you gotta play with the hand you are dealt with which in this case is the people that matched with you
|
On November 03 2014 15:27 Jaaaaasper wrote: If your support keeps pushing the lane you're kinda fucked unless you can get someone to pull. If not you just gotta deny when ever possible and only hit creeps for the last hit. That's what I was afraid off... Thanks for such a quick response though!
On November 03 2014 15:28 icystorage wrote: What Jasper said, sometimes you gotta play with the hand you are dealt with which in this case is the people that matched with you Okay, was thinking that might be the case, thanks!
|
On November 03 2014 15:29 Tannex wrote:Show nested quote +On November 03 2014 15:27 Jaaaaasper wrote: If your support keeps pushing the lane you're kinda fucked unless you can get someone to pull. If not you just gotta deny when ever possible and only hit creeps for the last hit. That's what I was afraid off... Thanks for such a quick response though! Show nested quote +On November 03 2014 15:28 icystorage wrote: What Jasper said, sometimes you gotta play with the hand you are dealt with which in this case is the people that matched with you Okay, was thinking that might be the case, thanks! Glad to help, its not like I have much else to do at this time of night.
|
|
On November 02 2014 01:38 trinxified wrote: Why inflation?
5 people always gain MMR, while the other 5 lose it. Always cancels out. Unless you're playing against players who haven't got calibrated yet.
Not to mention the abandons, where less people gain MMR over the ones who lose MMR. MMR inflation always happens to the top most players. Chess has a huge problem with this, because the majority of the time the best players win their matches due to lack of opponents that can beat them. Its been a problem for years, since the title of "Master" in chess is based on a fixed value.
Dota would have a similar problem, but would likely be less so due to the match making system and the fact there are 5 players. But inflation is still going to happen due to the top players having a tough time getting opponents at their skill level.
|
On November 03 2014 23:55 Plansix wrote:Show nested quote +On November 02 2014 01:38 trinxified wrote: Why inflation?
5 people always gain MMR, while the other 5 lose it. Always cancels out. Unless you're playing against players who haven't got calibrated yet.
Not to mention the abandons, where less people gain MMR over the ones who lose MMR. MMR inflation always happens to the top most players. Chess has a huge problem with this, because the majority of the time the best players win their matches due to lack of opponents that can beat them. Its been a problem for years, since the title of "Master" in chess is based on a fixed value. Dota would have a similar problem, but would likely be less so due to the match making system and the fact there are 5 players. But inflation is still going to happen due to the top players having a tough time getting opponents at their skill level. That's not technically inflation (which would be defined as a global flat increase in MMR) but rather an increase in the skew of the mmr distribution.
|
On November 04 2014 00:22 Sn0_Man wrote:Show nested quote +On November 03 2014 23:55 Plansix wrote:On November 02 2014 01:38 trinxified wrote: Why inflation?
5 people always gain MMR, while the other 5 lose it. Always cancels out. Unless you're playing against players who haven't got calibrated yet.
Not to mention the abandons, where less people gain MMR over the ones who lose MMR. MMR inflation always happens to the top most players. Chess has a huge problem with this, because the majority of the time the best players win their matches due to lack of opponents that can beat them. Its been a problem for years, since the title of "Master" in chess is based on a fixed value. Dota would have a similar problem, but would likely be less so due to the match making system and the fact there are 5 players. But inflation is still going to happen due to the top players having a tough time getting opponents at their skill level. That's not technically inflation (which would be defined as a global flat increase in MMR) but rather an increase in the skew of the mmr distribution. I don't fully understand the issue within chess, but an friend of mine is near master level and he said that everyone ELO in chess is slowly rising. It is likely not one issue that is causing it to rise, but a confluence of issues. It might just been the nature of any competitive rating system to have inflation in some way. Its not a huge deal as long as people are aware of it.
|
Inflation would only happen if more points are introduced to the pool compared to the amount of active players, but dota has something chess doesn't, a situation where there is lost more MMR than is gained. When someone abandons.
Which is why i don't personally believe there is a real inflation(and if there is, it's so miniscule it wont really be measurable), but again only valve knows this. I'd love to know the answer though.
|
I mean I really wasn't referring to the chess situation.
|
Sorry, I couldn't tell, I assumed you were, my bad. I am not sure if there is MMR inflation over all. Sure people could game the system and get like 7K, but that would level itself out.
|
Welp jugg is really good now.
|
Inflation in dota is mostly going to come from the smurf detection, imo.
In an actual zero-sum system (which dota has never really pretended to be), every player is started at the same point and climbs by taking points off other players. Dota's system, on the other hand, injects point whenever it boosts a new account, because the points that get a smurf to VH in four games do not come from the people in those games. Even though this happens before the account has a ranked MMR, the ranked MMR is created from that pool and is stll affected.
In the most obvious example, if a 5k+ player makes a smurf and boosts it to 5k-level by stomping his first dozen games, he creates around 3k "new" MMR. If he then hands the account to a 2k player who tanks it back down, that MMR is fed to the people the account loses to, which could conceivably cause inflation.
This doesn't work both ways because there is no corresponding catastrophic loss of MMR if a new account loses repeatedly - at worst, a tanked new account would drop around 1k while a boosted one can gain 3k+. Of course, there are opposing factors like the fact that abandoned winning games cause a net loss in MMR. I'm not sure that's a significant portion of games, but I could be wrong.
On the whole, I'm not convinced that there is significant inflation, but the mechanisms certainly exist. The easiest way to detect it would probably be to watch the cutoffs for High and Very High. Those seem to be percentile based, so if we saw the entrypoint to VH rising then that would be evidence for significant inflation. I don't think that's been observed yet.
|
|
|
|