We have a set of changes for hero rankings in the works, to be released after TI. The biggest change, which should solve for your concerns very well:
- Professional division will be replaced with a separate division that has both pro and very highly rated ranked players. Pro players will need to do more to earn their spots at the top of the rankings, and very highly skilled pub players stand a good chance at #1. High MMR ranked matches will be much more rewarding than they currently are.
THANKS alot! even if im no more top 5 after this change it feels much better than the current system! the problem is actually that not everyone can get into this devision.
your solution is a good step in the reight direction!
its is also unfairly skewed towards people who are like 4.2k but play 3,000 games with a single hero.
yeah, 2 important changes are
1. stop giving professional division such a big bonus
2. make the number of matches factor converge to a maximum amount of points otherwise its ridiculous
why do you put so much work in something that literally two or three person max gives a shit about ??
Remove winrate, KDA and number of total matches from the algorithm please. None of these things produces any reasonable or desired result. Dotabuff rankings so far have always been an abusing contest. The current one is at least useful, because you can find hero experts, i.e. people who spam the same hero over and over and steal their ideas. The one that came before this one was just a smurf-contest.
Way better stats that you should use instead are: Number of matches, Skill level of matches, KDA on win, KDA on loss. For ONLY ranked matches and ONLY matches within the last 3 months. People who didn't play at least 30 matches within that duration should not be able to qualify for the rankings AT ALL.
The current one is really bad in that aspect, allowing someone to qualify for a top position with barely 30 matches played, 29 of them more than 3 years ago. Horribly outdated stats should not be part of the statistic.
Another really stupid thing is that I can get 40-0 in normal bracket, then play a few games in very high and it will simply forget that the 40-0 was in a lower skill bracket and consider it the same as if it was in Diamond 10.
And professionals get on the top of the ranking for playing the hero in casual matches, that's also really stupid. "In pro games I only play the #1 role, but in my free time I play supports at 3k MMR bracket and inflate my KDA and Bracket with it".
everybody knows you can fake stats, thats why mods flag the accounts that do so
obviously they cant flag everybody but at least the most obvious ones (usually the ones in the tops)
And now I remember why people flamed smaug. Its not that he is bad (he isn't). its that he is so sure that he is the best OD that he sees any system that doesn't rank him as number 1 as flawed.
My thoughts are that it should only count games from the current patch, and be determined by MMR and win rate in those games. The minimum number of games played with the hero in the current patch to even get ranked should be fairly high. maybe let it count games from the last patch for the first month or so after a patch comes out.
Obviously it should be highest MMR -> Winrate -> Hero damage (withinhero) -> KDA (withinreason) -> then latest games as smaug said. perhaps not 3 months but last big patch so like 6 months.
I dont understand why pros should be treated diferently than high mmr pub players, because they play in SAME MATCHES, literally same ones, and are facing absolutely equal difficulties. Even with updates mentioned above by Jason, the difference will still stand, though, which seems to be not reasonable to me.
although I believe diamond division should be extended, cz there is a huge difference between lets say sf player with 1k games and 60% winrate at 5k and a guy with similar stats at 7k; they shouldnt both get same diamond-10 division.
High MMR ranked matches will be much more rewarding than they currently are.
fuck yes road to top 100 techies
I have a question
Is your 'division' in the hero rankings decided on your last games with any hero or is it decided with the specific hero of the hero rankings you are looking at?
So for instance let's say I have a KDA of 4 and 70 % win rate over a 100 games with witch doctor in normal skill games and then I never play with him again. Then a year later I am super awesome and playing really high ranked matches, and I've only played one witch doctor game in this high division, will it say I am in the diamond division or still the bronze division for witch doctor.
If it at that point it says I am in diamond division for WD then my stats are pretty inflated from my normal skill games and it's unfair?
@Triplesteal
But does it measure average difficulty of all your games or of only your witch doctor games?
Okay thanks!
I decided to not be an idiot an read the FAQ xD
Currently it seems that if you play a lot of games from the beginning with a hero (and especially at the beginning) it could negatively affect your ranking. As your average win rate/ average division/ average KDA etc are weighted down. SO if someone plays their first 1000 games with a specific hero but sucks in the first 800 and gets to a really high ranking near the end only it will take a lot of games of playing that hero to fix their rank.
Meanwhile they could be one of the best player out there especially as they spammed that hero.
For instance Fadi who I'm sure is one of the top 100 for IO won't get into the rankings? (unless I am mistaken)
Also for heroes like dazzle or omni I think healing should be a factor
As I said before, your winrate arguments are horribly flawed. You're literally claiming a 1.5k MMR player would be way way way better than a 7k player, because the 1.5k increased his MMR by 500 since calibration (for example with a 70% winrate) whereas the 7k player decreased his winrate after calibration (for example by 500 with a winrate of 30%).
I mean, you're flaming me for saying something that I never did, but your statements are so ridiculously wrong and yet you don't even notice it.
FACT is: Someone who sits at 5000 MMR IS A BETTER PLAYER than someone who sits at 4000 MMR, and someone who sits at 7k MMR IS A BETTER PLAYER than someone at 5000 MMR. If you honestly believe that two players with the exact same ranking who play against the exact same opponents with the exact same team mates and the exact same success are completely different in terms of skill, then you're fucking stupid.
FACT is also: If you order the hero rankings by winrate EVERY SINGLE RANKING will be a 1:1 mirror from the Dota 2 Leaderboards, so it's FUCKING USELESS.
Fact is also: the overall KDA is almost completely determined by winrate, so putting it into the equation is stupid as fuck. This is different from the KDA per win or per loss, which would be independent of someones winrate and thus would be obviously be way better.
But it looks like this forum is mainly filled with wannabes who can't accept the fact that they're wrong.
@TripleSteal-
You're wrong. It measures division for ALL your recent games. Else some people would be unable to get Professional division for a hero that can't be played in pro games (because of the CM restrictions). That's one of the flaws I was talking about already in my last post and you idiots conveniently ignored because you rather stick to hateposting instead of doing something that helps yourself.
@Crudude
SO if someone plays their first 1000 games with a specific hero but sucks in the first 800 and gets to a really high ranking near the end only it will take a lot of games of playing that hero to fix their rank.
Meanwhile they could be one of the best player out there especially as they spammed that hero.
Who would have guessed - oh wait it's actually me. Check my KDAs:
Last 3 months: 3.09 (on 168 matches)
Last 6 months: 3.10 (on 254 matches)
Last 12 months: 2.69 (on 812 matches)
All time: 2.52 (on 3920 matches)
Takes me only 15~30 owning matches to increase my KDA by 0.01. Nice. So fair.
I think the biggest issue with the population in this forum is that it seems nobody besides me has grasped the concept of MMR... :/
btw, Jason's statement is a copy pasta from a year ago I think. I know I heard pretty much the exact same thing back then. I wouldn't get my hopes up that they ever fix this system or even attempt to do it.
Yo Smaug
This is nothing to do with MMR, I am just trying to understand the hero ranking system
I read the FAQ and it does say the division for a specific hero can vary so unless the FAQ is wrong, triplesteal is right.
I kinda think now that the stats are a little weighted to your more recent games (yes? no?) so I think I was wrong anyway
I agree with your original post about 30 matches in 3 months, except I'd say more like a year as to play 30 games of one hero in 3 months is quite a lot. They could even have a yearly rankings separate to an all time ranking.
Anyway this doesn't affect me at all so I will stop posting in this thing xD
Indeed 30 games in 3 months on one hero is quite a lot, but you have to realize the purpose of the rankings is to find hero specialists, not just people who accidentally play the hero. When people watch those guys' games on the top list they want to be able to learn from them and not like I watched Black^ play OD and just face palm every few seconds because of such huge mistakes being made (such as him not being able to estimate cast times and cast ranges for example).
The FAQ is WRONG on that aspect, just like Tripplesteal. It can be proven VERY easily: Look at this heroes ranking: http://www.dotabuff.com/heroes/earth-spirit/players
Earth Spirit is not in captains mode, thus getting a professional bracket on him IS NOT POSSIBLE. Unless of course the bracket gets shared between all matches independent of the hero. It really has nothing to do with reading or believing any person. It's a simple FACT that it works like this.
some players are assigned to be there no matter what the mmr of their games is, this division is given to those ones playing in LAN finals
if you are pro 5, then you will be pro 5 with all heroes, but that doesnt work for non-pro divisions, which depend only on non-tournament games on this particular hero.
You have different divisions because you are playing the heroes infrequently. Whenever you play a hero your division gets updated to the one that your account is at.
发表评论请登录
this guy http://www.dotabuff.com/players/97353735 for example has 36 Pugna games and is Rank 8!!!! prof div 5 and NOT A SINGLE pugna game in a professional match. that makes absolutely no sense and professional is massivly overweighted anyway. just remove this crap. if they are considered to be so enormously better, they can easily go for top 10 with diamond 10, can't they?
many ppl will say "why u even care?". to answer this question quick. i i personally care but from my point of view it's absolutely fail to have professional devisions as explained above.
so dotabuff pls remove this crap and all will be happy :)