Page 3 of 6 FirstFirst 1 2 3 4 5 ... LastLast
Results 21 to 30 of 59

Thread: MMR very low after calibration, even after winning 9 out of 10 games

  1. #21
    Banned
    Join Date
    Sep 2011
    Posts
    12,296
    Quote Originally Posted by totemizer View Post
    stats for my calibration games:

    Code:
    game id	      Level	K	D	A	Gold	LH	DN	XPM	GPM	HD	HH	TD	KDA	    avg kda
    764769130	18	6	5	9	13.2k	46	0	558	404	5.2k	4k	969	10.20	    10.31
    766345987	19	4	7	16	13.7k	49	2	530	368	7.6k	3.5k	724	16.57	
    765660955	18	7	3	6	13.7k	77	0	595	461	4.7k	1.5k	1.4k	8.33	
    765590187	22	7	5	8	17.7k	106	0	559	374	5.7k	6k	988	9.40	
    764769130	15	3	3	12	11.6k	55	0	468	432	5.8k	2.4k	1.2k	13.00	
    755657997	11	1	6	0	7k	47	0	251	247	3.6k	0	73	0.17	
    750968575	18	3	12	10	15.1k	81	8	378	320	7k	2.3k	478	10.25	
    744488719	14	0	5	9	11.4k	44	2	290	314	4k	4.2k	666	9.00	
    742758046	22	7	6	16	21.1k	92	3	546	437	8.8k	5.3k	702	17.17	
    742603063	17	2	2	8	13k	66	2	448	367	5k	3.1k	277	9.00
    That hero damage. Is this a joke?
    Edit: nevermind, you were playing Omni. Dunno if the MM takes HD into consideration. Probably not.

  2. #22
    Basic Member Shmendrich's Avatar
    Join Date
    Oct 2012
    Posts
    2,579
    Quote Originally Posted by hoveringmover View Post
    It has already been established that they are doing this.
    (UN
    Omfg X) really? Do you have any official source who states this? That would be the dumbest thing DotA2 Dev team ever did....and they have done some pretty questionable ones X)

    Quote Originally Posted by Typhox View Post
    That hero damage. Is this a joke?
    Edit: nevermind, you were playing Omni. Dunno if the MM takes HD into consideration. Probably not.
    Hoveringmover says they ARE taking hero damage in to consideration, which would be like the stupidest thing ever.

    You can count things like: wards bought, hero heal, support items bought to measure a support...but how the fuck do you differentiate between a support that steals the kills from the carries, and one that doesnt gets any farm for leaving it for his carry? How do you measure a support that suicided to save a heavy carry?
    Last edited by Shmendrich; 07-10-2014 at 08:51 PM.
    Make a NON automated report system.
    I would rather wait 10 minutes for a balanced, fun and exciting match that will last 1+ hours. Than wait 10 seconds for a poorly balanced crapmatch that will end in 30- minutes
    Valve, How I'm I supposed to respect certain boundaries to prevent a mute if I dont know them because they are different for each and every person on each and every game I play?
    The only type of player that dislikes having his stats public are the players that suck and cheaters

  3. #23
    Basic Member
    Join Date
    Jan 2014
    Posts
    47
    According to some russian forums who are also famous for boosting accounts etc, hero damage is most important factor during calibration matches. Other factors are important too but hero damage can swing your MMR +/- 700 from your average. Before Valve patched it out, there was a console command that let you see your hidden MMR, and there are screens in those forums where player gained +300 MMR(Still hidden because calibration) in a game that was lost just because he had most HD. Also there were screens where people would get -100/200/300 MMR because they had low HD even though they won those games.

    Anyway these screens were done at the time when boosters were thinking out what exactly affects your MMR to figure out the system and exploit around it.
    Last edited by JPL; 07-10-2014 at 09:00 PM.

  4. #24
    Basic Member Shmendrich's Avatar
    Join Date
    Oct 2012
    Posts
    2,579
    Quote Originally Posted by JPL View Post
    According to some russian forums who are also famous for boosting accounts etc, hero damage is most important factor during calibration matches. Other factors are important too but hero damage can swing your MMR +/- 700 from your average. Before Valve patched it out, there was a console command that let you see your hidden MMR, and there are screens in those forums where player gained +300 MMR(Still hidden because calibration) in a game that was lost just because he had most HD. Also there were screens where people would get -100/200/300 MMR because they had low HD even though they won those games.

    Anyway these screens were done at the time when boosters were thinking out what exactly affects your MMR to figure out the system and exploit around it.
    Well... sadly I cant say I would be surprised if hey did something that obviously wrong...
    Make a NON automated report system.
    I would rather wait 10 minutes for a balanced, fun and exciting match that will last 1+ hours. Than wait 10 seconds for a poorly balanced crapmatch that will end in 30- minutes
    Valve, How I'm I supposed to respect certain boundaries to prevent a mute if I dont know them because they are different for each and every person on each and every game I play?
    The only type of player that dislikes having his stats public are the players that suck and cheaters

  5. #25
    Basic Member Inritus's Avatar
    Join Date
    Aug 2013
    Posts
    1,631
    Quote Originally Posted by totemizer View Post
    Sure, it takes account everything. But they said that it's the winrate what is most important. I am just curious what happened here. I accept that these are the rules, but I feel I was mislead. Why would I create a new account if my original account, no matter how shitty I play there will get better mmr anyway? Why would I pick omni and win every game with it, if the KDA matters more than the winrate? I feel I deserve some real answers for these.
    What? The only thing they've said about win-rate is that it doesn't matter at all.

    The reason you got such a low rating is that these calibration games only account for a small part of your rating. The majority comes from the pre-calibration games.

    And stop saying "Valve developers lied to you" just because you can't read and/or misunderstood their system.

  6. #26
    Basic Member
    Join Date
    Jan 2012
    Posts
    640
    Quote Originally Posted by totemizer View Post
    Code:
    game id	      Level	K	D	A	Gold	LH	DN	XPM	GPM	HD	HH	TD	KDA	    avg kda
    764769130	18	6	5	9	13.2k	46	0	558	404	5.2k	4k	969	10.20	    10.31
    766345987	19	4	7	16	13.7k	49	2	530	368	7.6k	3.5k	724	16.57	
    765660955	18	7	3	6	13.7k	77	0	595	461	4.7k	1.5k	1.4k	8.33	
    765590187	22	7	5	8	17.7k	106	0	559	374	5.7k	6k	988	9.40	
    764769130	15	3	3	12	11.6k	55	0	468	432	5.8k	2.4k	1.2k	13.00	
    755657997	11	1	6	0	7k	47	0	251	247	3.6k	0	73	0.17	
    750968575	18	3	12	10	15.1k	81	8	378	320	7k	2.3k	478	10.25	
    744488719	14	0	5	9	11.4k	44	2	290	314	4k	4.2k	666	9.00	
    742758046	22	7	6	16	21.1k	92	3	546	437	8.8k	5.3k	702	17.17	
    742603063	17	2	2	8	13k	66	2	448	367	5k	3.1k	277	9.00
    How is OP getting these KDA values? Can anyone enlighten me on his formula? Anyways, OP, I corrected the table for you:
    Code:
    game id	      Level	K	D	A	Gold	LH	DN	XPM	GPM	HD	HH	TD	KDA	    avg kda
    764769130	18	6	5	9	13.2k	46	0	558	404	5.2k	4k	969	3.00	    2.48
    766345987	19	4	7	16	13.7k	49	2	530	368	7.6k	3.5k	724	2.86	
    765660955	18	7	3	6	13.7k	77	0	595	461	4.7k	1.5k	1.4k	4.33	
    765590187	22	7	5	8	17.7k	106	0	559	374	5.7k	6k	988	3.00	
    764769130	15	3	3	12	11.6k	55	0	468	432	5.8k	2.4k	1.2k	5.00	
    755657997	11	1	6	0	7k	47	0	251	247	3.6k	0	73	0.17	
    750968575	18	3	12	10	15.1k	81	8	378	320	7k	2.3k	478	1.08	
    744488719	14	0	5	9	11.4k	44	2	290	314	4k	4.2k	666	1.80	
    742758046	22	7	6	16	21.1k	92	3	546	437	8.8k	5.3k	702	3.83	
    742603063	17	2	2	8	13k	66	2	448	367	5k	3.1k	277	5.00

  7. #27
    Basic Member
    Join Date
    Mar 2014
    Posts
    16
    Quote Originally Posted by Karyoplasma View Post
    How is OP getting these KDA values? Can anyone enlighten me on his formula? Anyways, OP, I corrected the table for you:
    thanks, tbh, i didn't even pay attention to that, so i made some terrible mistake there.

    Anyhow, everyone is telling that hero dmg is what's the most important. So I was completely misled and lied to by Valve.

    What's even worse, if this is true, then their shitty decision making are actually making it worth to drag all games as long as possible, and fountaindive as much as you can without dying.
    It also discourages people from playing as a team.
    It also discourages people to play more than just a handful of heroes.


    I believed them and I paid the price by wasting 150+ hours of my life just to find out that they are not just incompetent but they even lie to cover it.

    There is 2 common arguments related to this topic.

    1. they can't make it open, because people would abuse it.
    Well, apparently, people do abuse it, and others like me who would try to follow what they say, and win games, are laughed upon.
    As I am a programmer myself, I believe that it would be much better to have these rules shared with everyone, that way if people start abusing it, it could be fixed, openly. But I guess that would waste precious programmer time, which then they couldn't use for adding some random hero items.

    2. Computers can't differentiate between roles, and how people play those roles, that's why they must use the second best option like hero damage.
    I already said I don't believe this.
    They know who the good players are, there is a tournament with 90 of them going on right now where more than 10M will be given to most of them.
    But in the world there are tens of thousands of very good players. Those players are not all mid players or carries.
    If you look at the same numbers and instead of matching everything to some stomping carry stats, they could match the numbers to what roles everyone is playing.
    And it's not that hard to find out what role is someone playing. Like if I start with buying curier, wards, or buy wards later and place them(!!!!), buy sentries, that can be a hint.
    If I buy a sentry and sucessfully deward some ward that's not something a carry player would waste their time on, at least not very often.
    The type of hero can also be a hint.
    Then, if I stand around in the laning not trying to farm just deny and harass the enemy player, that's another hint that I might be a support.
    Then there is the pusher, where the tower damage and different itemization count. Then there is the ganker which roams a lot, buys dust.
    If someone sits down to analyze how the good players stats look like when playing different roles, I am pretty sure there would be countless many ways to judge what the noob player is trying to do. If there are 5 noobs playing all hardcarry then you can reward them based on that.

    What's good about this idea that it can be implemented in such a way that it dynamically changes the requirements as the game changes. Because the good player's numbers will look different after some major patches.

    What annoys me most, that supports are at least as important as carries in dota. This is common knowledge, no one (afaik) denies it. Yet Valve managed to implement an mmr system which actively discourages people to play the game as it is, and degraded into some stupid hack&slash like (not gonna name examples, but I am sure you all know some).


    SO yeah. I am frustrated because of the time I wasted, but it was my decision, I will not blame them any more than asking why they couldn't be open about it, at least don't say it's winrate when it's not. But the fact that they are consistently messing this up so much frustrates me, because to me Dota was always a community game. Sure, without Icefrog it would've not happened as it was, but there is one thing I know, that Icefrog always listened to the community and tried to be as open as it was possible (not about personal matter of course, that's irrelevant, but the game). Valve on the othere hand does the exact opposite, and it's not a good thing. They could leverage the community in more way than just syphon money out of us.

  8. #28
    Basic Member
    Join Date
    Mar 2012
    Posts
    251
    Wow you must be a real bad programmer if you say computers cant differentiate the roles.

    The roles are already known as you can see in the picking phase. You have symbols for carry, support, tank and so on on every heroe card.

    And now i show you a programming trick (vb.net) how to monitor different values depening on roll

    Select Case HeroeRole
    Case Carry
    Monitor HeroeDmg .....
    Case Support
    Monitor HealingDone, WardsPlaced .....
    Case Tank
    Monitor DmgTaken ...
    Case Pusher
    Monitor TowerDmg ...
    End Select


    Isnt that hard right :P

    /edit
    The only Reason i can think of why the didnt implement it is that this system might make it easier to gain mmr with certain roles than with others.
    For example its way easier to plant a ward than to do heroedmg

    But this is all only a matter of balancing it out correctly and Valve has engough data to do it
    Last edited by Raisti; 07-11-2014 at 03:09 AM.

  9. #29
    Basic Member Shmendrich's Avatar
    Join Date
    Oct 2012
    Posts
    2,579
    I didnt say it cant DIFERENTIATE I said it is pretty complicated to measure how effective is a hero regardless of his role.
    Quote Originally Posted by Raisti View Post
    Select Case HeroeRole
    Case Carry
    Monitor HeroeDmg .....
    Case Support
    Monitor HealingDone, WardsPlaced .....
    Case Tank
    Monitor DmgTaken ...
    Case Pusher
    Monitor TowerDmg ...
    End Select
    Omfg X) how many classes of Introductory programming did you take? 2? Java likely? X)
    Your example is so easy to prove wrong without even needing to go in to complicated numbers:

    Lets say you monitor mostly the hero damage of a carry:
    1 That is HIGHLY dependant on his support, you might get a carry with TONS of hero damage in a game all thanks to an ally
    2 Carryrambo: gets lots of kills, is never in teamfights, late game he has 12 deaths 0 kills but you can just never win the game because he doesnt defends or pushes with the team, he starts dying 4-5v1 and gg
    3 Carryfarm a carry that farms all early and mid game, is never in teamfights only until the end of the game might end up with less total damage than the heros who were pushing, defending and teamfighting.
    4 A carry warding is good when the supports are retard and dont do it and the other team is ganking a lot.

    Support Wards placed:
    1 location matters, and you cant go and say "Oh the good spots to ward are those" because it totally depends on the circumstances of the game (defensive/offensive wards, ursa/troll on enemy team, ganking wards, block spawn wards)
    2 counterwarding could potentially create people who buy 890890808089 wards to raise their scores.
    Healing:
    1 heroes like omni knight would have a huge score advantage over heros like shadow demon.

    Tank: damage taken
    X) Are you for FUCKING REAL X)????? please tell me you are trolling, you want to praise people who receive a lot of damage?
    According to your suggestion, a noob who loves playing with initiators like tide or bristle, that dies 389434789327 every game and receives 3497357349759834737895749k of damage should be in page 1 of pro players X)

    Pusher: tower damage.
    1 Heros who always (regardless of nothing) Keep pushing can and are often reason of lost games, they might get 6-7 towers a game and yet be the most useless player in the team, due to not defending or helping in teamfights

    And thats not even COUNTING heros who have multiple roles or that can act as other roles given its necessary X)...(Carry and support are total opposite roles, and yet heros like naga can do both, and quite nicely)

    What do you measure from a rhasta? He is an awesome pusher, a super support and can kill almost any hero with similar lvl and networth. (Yeah I've seen some games carried by rhastas)

    And for christ sake go read a basic guide about ranking and elo system, you cant use total values even in 1v1, much less in a 5v5 with millions of possible combinations of heros, items, lanes, roles and other values that affect the likely performance of a team (parties for example)
    Last edited by Shmendrich; 07-11-2014 at 03:23 AM.
    Make a NON automated report system.
    I would rather wait 10 minutes for a balanced, fun and exciting match that will last 1+ hours. Than wait 10 seconds for a poorly balanced crapmatch that will end in 30- minutes
    Valve, How I'm I supposed to respect certain boundaries to prevent a mute if I dont know them because they are different for each and every person on each and every game I play?
    The only type of player that dislikes having his stats public are the players that suck and cheaters

  10. #30
    Basic Member
    Join Date
    Mar 2012
    Posts
    251
    First of all i wasnt talking to you. i was answering to the poster above which was clearly indicated by not using quotes.
    Seems the lecture on quotes and posting you did in another thread (where i am on your side) would be also useful for you

    And the Code i posted WAS AN EXAMPLE. Where did i say that this is all ?
    Do you expect me to post 1k lines of code where all is considered ? Sorry i dont work at valve its their job to implement it
    Valve has engough data to find out the keyvalues for each role and to monitor them.

    Basicly you are saying with your post that HeroeDmg is the best way to monitor all Roles otherwise you wouldnt put in so much time to take my example apart

    But congratulations for taking a part an incomplete example i am proud of you
    Last edited by Raisti; 07-11-2014 at 03:27 AM.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •