We're going to be making another pass at hero rankings soon. I agree that pro players are too inflated. We will be eventually adding MMR data from Rating Buddy to polish up the top 10% of players.
We probably won't make it ranked only. However IF we feel confident in determining if a games was solo vs party ranked (big if here) we would give a bonus for solo ranked games.
I guess to follow up, we may make it so that ranked games are more important for high level rankings, but it's still possible for normal players who don't play ranked to get a hero ranking. They will just be ranked lower. So if you are striving to be in the top 100, playing ranked will be critical. But anyone can get a hero ranking if they play a hero frequently - they just won't make it to a great ranking without a high ranked MMR. Does that make sense?
There shouldn't be any professional tier at all. One of the biggest issues with your rating is that your divisions are completely decoupled from the game, so anyone can just get professional division and then gets a bonus for all of his heroes he ever played for literally no single reason.
Also putting Ranked and Unranked together is flawed, because the System works completely different. The only way to get a decently high winrate in ranked is by abusing or improving, but not by playing well. On the other hand, it doesn't seem to just make the games worse and worse for you the more you win in unranked, so over a very long time you can maintain a pretty high winrate in unranked games.
Furthermore, the system is considering centuries-old data. You really shouldn't treat my matches from 3 years ago the same as my current ones. Look, I have a 60% winrate on Lycan, but I haven't played the hero for almost 3 years now, still if I now would play a game and lose horribly, your system would still believe I was one of the best Lycan players.
Not to mention the drastic balance changes and even remakes the heroes got in that time. Someone who got 90% winrate on Tinker before his massive nerfs is simply unbeatable now.
^ I think it should only use matches in the last 3-6 months that way, you would get a fresh list of the current best players for the hero you've selected over that period as opposed to an all time data, this could also allow for better insight into who the best players are. I don't think I care about someone who's got 70% winrate 2 years ago on a hero then played a few games here and there to get back up there, nor do i care much about someone spamming a hero who has 60% over all but 40% in the last 3 months.
For example...
Someone who's started playing X hero and has had a lot of success in Y patch or period but who isn't a poster here so no way of knowing. Would be nice to see what they build and how they play, what they've discovered etc.
It will probably never be perfect and people will always have disapproval but I think using a more current dataset will help show current best players, and allow for players (like smaug) revisiting favorites to effectively start anew each patch.
Just my opinion.
I think what Smaug is saying is that instead of basing rankings off lifetime stats, do it in a shorter window period.
For example by patch. This would account for the current meta. For example, I'd be able to see the currently skilled jug or void players and not players who spammed them before they were nerfed.
It's should not be simply short window, it should be shorter window with full effect, and then having less importance.
If one was a god at "enter hero name" 1 year ago, and stopped to play it, it doesn't mean that suddently they suck at it. It would not be normal that they need to play again 50-100 games at this hero to be shown again in the top, if one play like 10-20 games he would most likely recover his old skill at it
erm matrice you are totally fucking right m8
mmr= overrated junk
pro players=pfft kill me
meta+ patch shifts= LMFAO
i can literally go the same stupid yet cheesy and scrub obliterating TB build that I invented by and for myself after watching a vid by restocker AFTER reading about my schoolmate DrZack's posts on playdota and reddit about your TB ways, IN THIS PATCH
and I'd feel zero difference from then till now
how is that possible?
HOW is that possible?
ignoring how awesome I may be at the general understanding of the meta itself OR at the hero regardless of patch I think it's fair to say that what you said is true, it is possible for an old player of a hero to recover at least MUCH faster than those who just HURDUR play a hero for the sake of it being the flavorofmonth
also FUCK YOU icefrog, yall know what i'm talking about tinker tb pickers
fucking anti fun patch changes FOR SOLO PLAYERS
ignored tyvm random ledditard-ish interweb trollfuck post TI3 peasant convert stacking MMR whoring fuckface
^also
30% winrate all time ranked MM with sf 2ish kda 400 ish gxpm
LMFAO
senpai teach me how to sf
http://www.dotabuff.com/matches/1088913780
nice quad aquila build next level uwot m8 8/8 serious l8git m8
The best would be:
- remove Prof-division
- count only ranked games (for balance reasons), but don't calc in the MMR-value
- count only the games of a period (maybe 6 months as mentioned above) | (game limit down to 25?)
- keep the rest of the algorithm
these steps should make the ranking way more accurate and reprsent the players ACTUAL skill much better by considering only ranked matches.
btw im actually pugna #7 :P
I never play Kunkka but actually quite a fun hero http://www.dotabuff.com/matches/1400659213 I don't think I will ever make top but he comes online so early.
Obscure heroes seem to easier to get, I think the overpopulation of "Professional" makes it hard to get into common hero ratings.
Weird seeing someone main pugna :)
^ye, thats right. getting to some SF/storm top is nearly impossible since actually 100th place has as much score as top5 in other heroes' tops.
^where are u from, dude? i barely understand ur eng and dont remember meeting ppl who speak it this way recently.
კომენტარის დასაწერად გაიარეთ ავტორიზაცია.
Remove this shiet please, every1 plays the same matchmaking.