EPT2.3k means you're pretty much the average DOTA2 player.
Simple Questions, Simple Answers - Page 1046
| Forum Index > Dota 2 Strategy |
|
Birdie
New Zealand4438 Posts
2.3k means you're pretty much the average DOTA2 player. | ||
BluemoonSC
SoCal8910 Posts
so not only do you have to sign up for dotabuff, you also have to opt in on their mmr tracker. slight bias in the data set, but i'm willing to bet you're average. also keep in mind that the skill gap between 3-4k is massive, 4-5k is massive still and 5-6k is greater, yet  | ||
|
Belisarius
Australia6233 Posts
The H/VH cutoffs were something like top 15%/5%, and those correspond to around 3k/3.8k. That's probably the most reliable data around, and it isn't much. There's also the MMR breakdown in the Ranked reveal blogpost, but I suspect that's pretty outdated by now. On February 24 2015 06:29 Birdie wrote: One of Dotabuff's staff mentioned that the median is about 2250 MMR if I recall correctly, with only about 20-25% of games in the High Skill and Very High Skill range. Also source? | ||
|
Logo
United States7542 Posts
| ||
Excalibur_Z
United States12248 Posts
(Note: Actually I have to amend that, because I remember asking my old Valve friend at the time where the H/VH markers were in terms of rating and percentiles and he didn't know. So that may not actually be right. I remember phantasmal's post that said it was closer to 80/16/4 so that's probably more correct. The stdev stuff was right though.) Now that ranked matchmaking is in the game and carries a separate MMR, you almost have to clean-slate it. Players tend to play differently, and some players don't play ranked at all. Only Valve knows just how different the ranked skill curve looks compared to the unranked one, but I think Dotabuff is still going by the old unranked breakpoints. To speculate just a little bit: when you start out in Ranked, you play 10 calibration matches which affect your seed MMR more greatly than other games and have certain skill analytics measurements (based on hero damage) which the community calls "smurf detection". This is how fountain-sitting Zeus-ult-spammers were able to jump up to 6k after their first 10 games. If Valve used their unranked skill curve as a prior distribution for their ranked skill curve as a function of hero damage, then their algorithm could conclude stuff like "this guy got 100k hero damage in this game for this game length, and only the top 1% get that high, therefore his starting MMR should be around the top 1% which is 5500." So that could work just to scatter people around at the start, and if calibration is wrong they'll eventually drift toward where they actually belong. For that reason though I don't think you can definitively say that X MMR equals X percentile for Ranked unless you excluded everyone with less than, say, 200 ranked games played. If I played 10 matches and got 5500 that's one thing, but if I played 200 matches and still have 5500, that means I'm probably pulling my weight and earning that rating. | ||
|
Unleashing
Denmark14978 Posts
| ||
|
Belisarius
Australia6233 Posts
On February 24 2015 07:02 Excalibur_Z wrote: So years ago, before ranked was added, the mean was established at 3000 with a standard deviation of 350, and High and Very High were +1σ and +2σ, or 3350 and 3700. +1σ is the 68th percentile and +2σ is the 95th percentile. Need a source for that too. There is so much misinformation around on this topic, and a mean of 3k is very different to the actual data that Valve released. http://blog.dota2.com/2013/12/matchmaking/ 5% 1100 10% 1500 25% 2000 50% 2250 75% 2731 90% 3200 95% 3900 99% 4100 | ||
|
Excalibur_Z
United States12248 Posts
| ||
|
Belisarius
Australia6233 Posts
The familiarity question thing still happens btw. I'm not sure where it seeds you exactly but the "I am a dota god" option usually puts your first game in high, so 3k-3.5k sounds about right. | ||
|
FHDH
United States7023 Posts
| ||
|
Excalibur_Z
United States12248 Posts
| ||
|
FFGenerations
7088 Posts
| ||
|
BluemoonSC
SoCal8910 Posts
maybe we should make another thread if we wanna talk about what we assume the avg mmr is, etc. | ||
|
Belisarius
Australia6233 Posts
| ||
|
FFGenerations
7088 Posts
http://www.dotabuff.com/players/114635809/trends?date=3month&hero=&skill_bracket=&lobby_type=ranked_matchmaking&game_mode=®ion=&faction=&metric=gold edit: does this graph mean that my current average kda is 4.18 ? so i can say "i play 2.7k mmr dota with a kda of 4.18" ? http://www.dotabuff.com/players/114635809/trends?date=6month&hero=&skill_bracket=&lobby_type=ranked_matchmaking&game_mode=®ion=&faction=&metric=kda ps: i emailed dotabuff to ask them about adding a filter to show solo ranked matches (instead of mixing it with matches where you are in a party...because i dont play seriously in those...) and they replied saying they'll look into it. maybe if you feel the same way go ahead and email them too to show support.....? | ||
|
l3loodraven
2753 Posts
| ||
|
FFGenerations
7088 Posts
On February 24 2015 11:20 l3loodraven wrote: I'm sure the people running dotabuff will want to be making more money before they add any substantial new free features. i would subscribe instantly (to support the site) if they added that functionality because its the only functionality i really care about (my solo ranked mmr). mixing it with my party mmr makes it useless to me... i mean , dota mmr only records 2 things: ranked and solo mmr, so you would expect it to be a priority feature. they have a "substantial" range of free features already (even trend graphs!) | ||
|
FHDH
United States7023 Posts
And for those who don't know they are only able to do this if you accept a friend request from a bot account. As a DB subscriber I have to say the TrueSight parses are pretty excellent. I actually find the trends feature to be one of the least useful aspects because I am a data whore and line charts have to be pretty excellent to be useful. Being able to go deep on individual games though? Good shit. | ||
|
FFGenerations
7088 Posts
truesight would not be useful to me, i just want to track my solo ranked winrate basically and thats it. since i play maybe 50% of my games with a friend its impossible to tell what my solo winrate/stats is it is probably impossible to implement currently because otherwise it would have been..... their message was "We'll look into adding a filter for solo and party. Thanks for your input!" so fingers crossed | ||
|
Kotreb
Croatia1392 Posts
i tried searching for a year beast thread but i only found one link which was closed :/ | ||
| ||
