Its because rankings are innacurate. For example probably you will get more points if you lose game with me because I got higher mmr and opponents is stronger,than winning in solo queue or with other 3k players.
that happens because the most important factor in hero rankings is your division, so even if you lose a match you can get a higher score if it was a higher division game
on the other side if you play with noob friends you'll lose a fuckton of points :^)
simple. you move up and down the rankings when playing ranked games - and so does others.
It could involve a number of factors. My solo unranked MMR is about 1k higher than ranked. I get Very High on solo unranked games, but Normal on solo ranked games. Maybe this confused the ranking of heroes depending on their distribution of games played at widely varying MMRs. And as Bogi points out party MMR for a game could vary even more widely. I've been in parties with players 3k above me.
Ok clearly the impact of the most recent game is strong.
http://www.dotabuff.com/matches/1489439347
I played a Normal game on Veno and won with 6-8-28... lost 300 points ranking on the one game. lol. So its very much a "what have you done lately" score.
,,,,,,,,,,,,,,,,,,,,,,,,
Yeah, it was a very easy game. 3.2k
That was a rather unusual build for venomancer but my items worked out well for the situations and capabilities of that game. I would not change them. I finally managed to get my MMR low enough to win 40% on Jakiro and get a score (2468) 2nd lowest of all ranked heroes so far. I'm sure drow will be even worse. I might have to drop below 3k to make drow win over 40%.
He probably just accidentally double posted.
I recently got back on dotabuff plus, since I now have enough time and a decent internet setup to play dota 2 somewhat regularly. I really like the Truesight features and the comparison sorting that is available.
I was hoping to see rankings for all my heroes but I have not actually played enough games to make them meaningful for most heroes. I've got 24 ranked heroes, but I would need a minimum of another 1151 games to get all off them ranked and then a number of them would have to be improved considerably. I've got 26 heroes with winrates below 40%. It would be hard to get them all up to 40% winrate and current (played within a month).
I see a very wide variation in hero scores based on which type of game I played recently (over 2k spread). And if I play a hero that was performing well in Very High games in a Normal game it drops the hero down hundreds of points even if I win. Conversely if I play a hero that was failing even in Normal level games and play it in a Very High game, and lose it jumps up hundreds of points despite the fact that I lost - even achieving terrible stats in a Very High game is worth far more than crushing a Normal level game. That in and of itself would be fine, but the score seems too heavily weighted on the most recent game.
This doesn't make a lot of sense. For instance, https://www.dotabuff.com/players/103176226/matches?date=&hero=luna&skill_bracket=very_high_skill&lobby_type=&game_mode=®ion=&faction=&duration=&enjoyment=any
Luna is 5-1 in Very High games, with KDA of average of 4.66. But since my most recent Luna games were played in Normal the hero score is one of my lowest (2,362)... I know I play Luna well enough to gain MMR even when I was at 4k level. So the system doesn't seem to deal with variable MMR very well.
I wonder if the same problem might exist for the highscores as well. It seems that it would be possible to rack up great stats on a hero in the 4k to 5k range, then play just one game of it at 6.5k and jump to the top of the highscore despite failing on the hero at 6.5k level. It seems too heavily weighted on the current MMR of the player rather than reflecting the MMR at which the stats for the hero were achieved. If this game https://www.dotabuff.com/matches/394222535 , or this one, https://www.dotabuff.com/matches/374503366 were my most recent lich game it would be rated 2k higher than it is just because those games happen to have verified pro players in them - if the most recent game is Diamond 8, but my typical play is Platinum 3...it doesn't make any sense for it to jump up into Diamond if I que with one friend and then down into Silver when I que with a different friend.
Instead each game should contribute proportionally to the overall record of the hero. So if just one game in 30 is Diamond level and the other 29 games are in Gold, and the Diamond level game was lost and the hero fed... it should not have a huge impact on raising the average. Or conversely if one game was in Bronze it should not drag the Gold level hero down, especially if you were Beyond Godlike in the Bronze level game.
I think by now everyone is aware that dotabuff's rankings are about as legit as Dota Skill Rating (DSR) a week before ranked came out.
K/D also doesn't mean shit.
The description says they are still in Beta testing. I assumed they were still making changes trying to improve the rating system.
i was top50 on every single hero then they remade the stats and gave "PROFESSIONAL DIVISON" players a couple thousand points for no reason and now im like #20 luna and thats it
Veuillez vous connecter pour publier un commentaire.
https://www.dotabuff.com/pages/rankings#player-skill
I have noticed that my rating on a particular hero may rise several hundred points on a single game and the division jump from maybe Gold 6 to Platinum 1 in a single lost game. Also a number of times I have lost ranking points despite winning a game.
Why does this happen? Is it weighing recent games more heavily? And how do I gain a lot of points losing a game, but lose points wining a game? Did a spectacular win from the past fall off the back end of the "recent games" in the calculation causing me to lose points when it updated?
My Drow and Jakiro are quite bad, with winrates below 40%. They both have over 30 games, but do not have a ranking. The note says "Play more matches to increase your winrate and get ranked." Do I have to achieve a certain winrate for the score to count?