General DiscussionSome usual content does not matter on MMR or skill brackets

The way you express yourself is a bit... weird. Smells like google-translate.

The core factor to get to higher skill bracket is winning consequently, even when the average mmr of games increases. It doesnt directly correlate with winrate though, cause it depends on the level of games you are plsying in. Lets say you can have 60% winrate but be Normal Skill player cause you didnt manage to win any high skill games you were put to, and ur winrate is high only cz you play at 2k mmr instead of 2.8k that you deserve.
Or you can have 50% winrate and be VHs bracket playa, cause you win 50% of VHS games you have, and therefore do not drop to lower bracket.

Sorry, I speak 4 languages but no one in perfect stage. =P
When I start this acc, I got like 70% of wins plus kda > 5.5, and still normal brackets.
There is many, really many of players in VH brackets, with kind a new acc having less victories and less kda as well.
I thinking to do some statistics about this but I have no time. =/

I did an experiment trying to see how high can i calibrate playing only supports, and i noticed things that could help your case.

1. You choose your previous experience in the genre, there are three options, likely giving you 1000, 2000 or 3000 starting MMR based on what you chose.

2. First ten unranked games serve the same purpose as the first ten ranked games (ranked calibration). I believe it gives you 100 MMR for each win, and takes away 100 MMR for each loss. Note that you will be matched with people who have similiar amount of games as you. That makes these first few games very unstable in terms of players' MMR. That's why sometimes first game for some players ends up being High (or sometimes Very High, but that's extremely rare) instead of Normal.

3. 100+ unranked games until lvl 13 work the same way as ranked games (+-25 MMR). You will still be matched with people with similar amount of games for a few more matches. After that, being matched with people with over 2000 games is entirely possible.

4. At lvl 13, ten ranked calibration games (+-100 MMR)

5. You get your ranked MMR which cannot exceed a certain MMR (4.5k or maybe 5k).

It is possible that there is some extreme version of Anti-Smurf Detection, that propels you up or down by significant amount of MMR, but i doubt it (or at least, there isn't one anymore after Zeus spamming). Basically, if you're a 5k player starting a smurf, you shouldn't have a problem winning 10/10 first games, which would instantly get you 4k MMR. Next 100 games, plus 10 calibration games, will further narrow down your MMR, and likely will put you exactly where you belong.

Although i started this account in March 2014, and calibrated only a few days ago (i don't have time to play too much), it is possible that Valve changed the system a few times during that stretch, so my conclusion might not be 100% true, but i believe it's not very far from it.

regarding ur point 1, if one wants to land in the highest mmr possible, they should obviously always pick the option stating they have a lot of previous experience in the game/genre right?

my 1st account had 66-70% win rate and calibrated at 2,8k

win rate means nothing. all that matters is in what skill lvl you play

@Santana

Yes. But still, one can pick "No previous experience" option, and still end up in 4k+ after calibration. Obviously, it's a more painful road.

@Caesar

1. You might have picked that worst starting option, and started at a very low MMR.
2. You might have performed very poorly in the first ten games, which resulted in a big MMR drop.
3. I don't know if you played exclusively solo on that 1st account, but if you played a lot of matches in a party, than obviously Solo MMR will remain unchanged, even if your overall winrate is increasing.

I found that case in a quick search, http://www.dotabuff.com/players/214794571
They have on wins and still getting highs brackets.

@ Tchub Dju

Have in mind, just like it's hard for 6k+ players to get matched with other 6k people (usually 4k's are involved), similar case is here. Not many players with low amount of games (usually single-digit) are there when a player with low amount of games tries to find a match. MMR of players in those games vary significantly. I wouldn't be surprised if MMR difference between players in those games surpasses 1000. That's why one game it's Normal, next Very High etc.

Also, there is a possibility i thought was realistic some time ago, when a system during calibration, simply throws a player from one MMR bracket to the other, eventually finding a suitable spot for him. Win rate wouldn't be a decisive factor in this case, and a game would have to look at other stats (APM among other things maybe...). Game might still work this way, who knows... Or maybe some combination of everything i wrote. As i said, i only used this account in attempt to find out the algorithm. Far more evidence (accounts) would be needed for something more reliable.

Valve must have some sort of ip-related smurf detection in place. I've seen smurfs mark "never played a moba before" and still get put into VHS while solo-queing on their first smurf game.