Skill bracket data is derivative of DBR. We do not assume that old DBR data is "correct", but we do use it (along with a confidence value based on completion, sample and age) as a component of our analysis to guess at the skill of a match. If the analysis doesn't have a high enough confidence it throws it out entirely and the match doesn't count toward a skill bracket.
Of course, this system grows towards having no data over time. We're in the process of replacing it.
Hi,
Jason could you also reveal the information about the MMR leak in october 2012, how that was possible ? and what are valve's MMR calculation factors? does kda matter?
thanks.
Sorry, I can't speak to the details of Valve's systems. We don't know how they calculate MMR.
Hello Jason, is it possible to make Plus users see their current DBR, with the info that it probably is not correct?
Yes, I know that, but you guys surely kept calculating new values, right? Could you make those possible to be viewable?
i highly doubt about it u can make dbr out of all the dotabuff users that has stats sharing enabled
Which are not as many users as you might think.
I guess the dotabuff devs know more about the feasibility of calculating a meaningful rating than us ;-)
My bet is bronze players (<40%) don't have stats sharing enabled. How can you calculate the DBR of a 60% player when there are 0 players in 0-40% range?
Please sign in to post comments.
Even though the dotabuff rating was phased out a while ago, you guys still seem to be using a few stats that rely on knowing it.
Player ratings for specific heroes are heavily dependent on it. We can also sort hero stats based on the bracket, which requires some kind of rating. Bronze through diamond, not just normal to very high.
Is it just a rougher estimation of the old dotabuff rating made with assumptions about anonymous players? Or does it still use players 6 month old DBR as current (that would be flat out wrong and dishonest to do...)