yeh they are full of shit when it comes to describing things they add/use in dota 2 just like report system they make it sound like they had some super complex big thing lol and it was some simple shit alghoritm that got exposed by f2p on devs forum/reddit
so if the mmr ranking is based on kda, zues and specter players will be the top mmr, people that play chen/treant/omni will be the lowest mmr
if mmr is based on damage delt, everyone will pick carries
if mmr is based on wards bought, even carries will buy ward
win/lost is the most fair way to calculate it. u should be happy valve even implemented a mmr system, they didnt have to do it.
Hi Commend,
NOTE: Most of my post depends on the speculation that Valve follows a variant of a Trueskill system. And my understanding of the Trueskill system from Jeff Moser's site: http://www.moserware.com/2010/03/computing-your-skill.html
"In fact it goes down -27, more than it has in some other games where I didn't play as well. Listen Valve, don't lie to your devoted audience."
Valve says "Win/loss is the primary criteria used to update MMR, but individual performance also plays a role, especially when our UNCERTAINTY ABOUT YOUR MMR IS HIGH."
If Valve is using a variant of the Trueskill system, then you lose more points or win more points depending on how much they failed to predict your win or loss. If the uncertainty is high (like when you start playing or start playing the calibrating ranked matches), I think there is a chance you gain points while you lose.
I think your Mean MMR is increased for wins and your uncertainty is reduced for correct predictions. The shown MMR is a conservative estimate that can go up after a loss. That is if the uncertainty has'nt reduced much compared to the increase in mean. The score assigned to your team is most likely based on how badly they estimated the closeness of the match(as valve calls it 'quality metric'). So in your case, perhaps your team was supposed to have higher 'quality metric'. Sadly there is no way to test my theory...
@lulu aka retard
They can compare your stats to the average for that hero (possibly by % of totals, further divided by brackets, game length and win/loss too).
wow, that was hard
edit: That isn't to say that its perfect, but what you are arguing is not a problem in the slightest
Yeah casuals found the way to ranked, My past 14 games have had a dc now, its a shame.
@hobbit
^whats your MMR? judging from your recent games its not too high, its not surprising to find 1-2 retards drifting into your matches even at 4500-5000 MMR levels, and they keep appearing for higher players, just less frequently
@Terrible:
"They can compare your stats to the average for that hero (possibly by % of totals, further divided by brackets, game length and win/loss too)."
Thats not a good idea since heroes can be played in different roles and you will be penalized/rewarded for not following the trend.
For example, a support alch will have low gpm/kda, while carry alch will have exactly opposite (if played correctly). On the other hand, the actual average of those two would be a mediocre alch who perhaps failed both as a carry and a support :)
I noticed it takes less negative points when you are playing AM and losing... Once i got -20 with AM ( http://dotabuff.com/matches/432551252 ) and lately i got -22 ( http://dotabuff.com/matches/439783242 ) and I did shit, and averange point loss is like -25 for me.
Look, I have had almost the same situation like you OP but I just had a better hero to be good with http://dotabuff.com/matches/439834345 tinker and furion 2 pushers, my team does shitty to def and I, 14/1 and I have got -25/26 for that game too. Yea I was trying to go back as often as I can while trying to push but when I come no one goes for the kills and most often there's no dust to follow.
And I was the one to blame after all lol, OD and magnus was butthurt about me since the very begining for some reason, I think getting soul ring (good for early harras and kill) and threads wasnt liked by them - magnus was saying some shit I started 1 build and went with the other while I just was getting what I found usefull.
I think it is pretty obvious what needs to happen with the currently broken MMR system. Obviously winning and losing needs to be a factor, if it isn't people will only look to farm kills and winning will become secondary. Even if most people still care about winning there will always be one asshole looking to pad his MMR and fucking up the game for everyone else.
With that being said personal skill needs to play more of a factor as this thread shows. You shouldn't be going 30-1 and losing a ton of points.
I think the only way to truly do that is to COMPARE how you did to your teammates. Now obviously this is complicated as everyone plays different roles, but valuing assists with kills is important. Overall hero damage dealt, tower damage done, healing done, your KDA ect.
I'm not saying you should gain points even if you lose, (or lose points if you win but played like shit). But there should be averages that are made for your team (and within that, MMR Bracket ranges) and then your stats compared to that. If you played drastically better than your team but still lost, and still contributed to the game (tower damage, assists), you should be penalized much less than those who fed. Maybe only losing 10 points. Obviously the numbers would need to be figured out.
TL;DR Make an average for your team, compare it to that, make percentiles of better and worse score points based on which percentile you fall into, factor that very strongly into the win/loss.
"Overall hero damage dealt, tower damage done, healing done, your KDA ect.
I'm not saying you should gain points even if you lose, (or lose points if you win but played like shit). But there should be averages that are made for your team (and within that, MMR Bracket ranges) and then your stats compared to that."
As Sampson said, your idea favors carries. It will hit supports harder. Considering they might die first in mid-late game team fights, they will have lesser assists and hero damage. Disabling heroes is not counted in these stats and a lot of supports disable heroes well. Further giving your team vision by constantly buying wards/ smoke and buying mek/urn also cannot be calculated in a foolproof way.
Also it might stereotype heroes into playing a high stat-obsessed role and constrain creative usages of the hero that gives rise to new metas. I remember jerax saying that he likes playing Faceless void as support and he has high winrate with support Void! ^_^
Supports can still get assists, and they have things such as healing which carries wouldn't get which you could value. There would have to be another intangible that you could measure, based on support attributes. But obviously stuff that you couldn't exploit. That's the challenge. I still think it could be done though. And regardless, the principle being team comparison. Regardless of role, if you play well in comparison to your teammates you should be rewarded.
its too easy to exploit things like kda hero heals tower dmg etc. The only thing that matters is the win or loss
But the win or lose doesn't say how well you played. You can play out of your mind and have a team of feeders. Hence why it should be done IN COMPARISON. I'm not saying winning shouldn't matter or even not be the main factor, however it shouldn't be the ONLY factor. How well you played next to your team, is important. Also take into account a bracket of MMR (say 3500-4000). The average kda, healing, damage done, game length ect. Take the average for each hero in that bracket, and also use that. Say the average KDA of Lich was 2.5 and you got 3. You scored in a certain percentage of players so you get rewarded, also base this off damage, healing ect.
It should all be based off points attributed to IQR. Quartile ranges based off percentages from the mean. Both in a specific game and in your general skill bracket.
The system should not be based around who does what better than whom, because, as stated, these types of systems can be exploited. If they tried to make a system around tracking support, people would spam wards/smoke. If they tried to implement damage tracking, people would find ways to pad the numbers.
I think what InVerum is trying to say, and something that most players can agree on, is that the current system is simply unacceptable. Yes, DotA is a team game, but it's simply not correct that individual player skill should be ignored, because otherwise, what's the point of tracking individual scores? Any system that tries to track in-game actions is unrealistic; this much is obvious, especially with the rapidly-evolving meta ("Which hero is a support? If I pick Naga, am I being tracked as a carry, or a support?").
However, it's relatively easy to look at numbers and judge individual player performance. The game tracks stats like hero damage, tower damage, healing, etc., so using the existing metrics would not be difficult. In my opinion, the best way to avoid abuse should be to develop a system that looks for large discrepancies between players, but otherwise maintain the status quo of win, gain points, lose, lose points. For example, for the winning team, the system should check for contribution. Did everyone do some hero damage? Was there one person who literally had 5 K/A, while the rest of the team had 20+? Did this individual have a high GPM/CS? Or for the losing team, was there one person who went highly positive, while the rest of the team went negative?
The system should not be based around "how good you did", but rather, "how good you did not do". It should be based around tracking NEGATIVE behaviours, such as jungling the entire game and not contributing to pushes/kills, or going negative in a losing game when your teammates are positive. This way, it would encourage positive behaviour, thus the "reward" is the normal point gain, and would (hopefully) be subject to less abuse.
No system will ever be perfect, and I'm not saying that this would be, but I'd bet money on the fact that we've all had that game where we go positive, do the most hero/tower damage (even as a support), and still lose because of X/Y/Z reason. In what universe is it appropriate for one to lose points if they genuinely performed well, especially in adverse conditions?
^ This is probably the best option. If you win you get certain points, if you lose, minus certain points. However, if you win, and are drastically below the average for your team, you don't get as many points. And if you lose, but score drastically higher than your team, not penalized as many points. This would essentially make exploiting pointless. Because you still get set points based off the win.
So you can only be potentially penalized on a win if you didn't contribute, and only potentially rewarded on a loss if you did drastically better. Obviously would have to change the limits for each. Have to do REALLY bad to be penalized on a win (again still getting points). And only have to to do quite a bit better to be rewarded on a loss (again, still losing, but maybe only 10 instead of 25).
accept it guys the only way to the top is stack and stomp...else just play with baddies and feel bad..this shit is just gay...but then thats how it is...also when u solo que..they shld no pool u with ppl in party..but they do it anyway..so it kinda sucks..even a 3 stack against solo players makes it bad
Lost 20 rating from this match
http://dotabuff.com/matches/425177409
Calibration phase, possibly due to low hero damage. However, such factors don't seem to have any effect outside of calibration phase anymore, except perhaps to change your uncertainty.
I'm not even complaining, don't see why all these tards bother. Never played any Elo based game before?
Outside of calibration, changes in rating mostly depends on average rating of both teams. Against a team with higher rating you gain more, and against a team with lower rating you gain less. You are more likely to get a big rating difference when queuing as 5 stacks as the allowances tend to be bigger; i.e. the game likes to match five-stack against five-stack where chances are there is a larger average MMR difference than with small parties and solo-queuers.
"what's the point of tracking individual scores? "
Like relentless stated a lot of times here i think valve thinks this is a team game, that its probably the reason they didnt want to keep track of individual perfomance before. Just imposible to create a perfect ranking system for this game, the actual one has a lot of fails, but any other idea you guys have will probably have even MORE fails.
as Dork stated while the system is not perfect it is the best that can be obtained from dota pubs. Once you start giving too much credit for individual performance you pull away from the team aspect that dota is based on.
For example as much as alot of the pro gamers joke about "space created" when they get 5man gangbanged it is a legitimate strat to work on. (The whole opponent team is focused on getting killing on your CM while your AM gets time to get his core items).
It's funny because I posted the same exact thing in the steam community here:
http://steamcommunity.com/app/570/discussions/0/648814842910994709/#p1
This guy started attacking me saying that computers would have to be sentient for a system that's better than win loss. Then some other guy talks about how him and some friends in fact created a system that works better than win loss and used it effectively and he gets attacked by the same guy. Thanks dotabuff community for being more open-minded and having a productive discussion. I think this is an important issue that Valve has to address. There needs to be more transparency in the current system, and they need to be open to suggestions for improving it. As we see above, unless more complicated variables are introduced, simple variables such as wards placed, k/d ratio, etc. can and will be abused. My hope is that a system is one day created where the abuse is limited by additional variables to the point where the matchmaking becomes more accurate than win loss, because let's be honest, win loss is too simple of a variable to determine skill when you're 20% of your team.
u stomped the game, which shows that the other team was probably worse than urs, but even beeing more skilled players YOU LOST THE GAME. Can't u see it? The other team strat was better than yours, so they deserve to win, and about the points u lost, i think it should have been more, since the other team seems to be a lot worse than urs.
Dota is not about killing or farming, it is about breaking towers and destroying enemy base. So don't try to convince me that good scores makes u a good player. Breaking enemy base makes you a good player, and for that u deserve a nice MMR.
edit: of course the players of each team should receive diferente points, considering their solo efforts for the win/loss of each team, but no doubt the main thing to be analysed is: "who won the game?"
''Dota is not about killing or farming, it is about breaking towers and destroying enemy base.''
This is main problem,people don't watch pro teams or any other streams,they don't have a clue what is the main target in game,they dive and die,they loose even having 10+ kills because of ''just one more item pls'',and that item is not BKB on hard carry,they chase one kill with 4 members and loose 2 towers in 1min etc etc.....
Totally agree with you Coutones.
Please sign in to post comments.
"Dota 2 uses standard techniques to quantify and track player skill. We assign each player an MMR, which is a summary metric that quantifies your skill at Dota 2. After each match, we update your MMR based on what happened in that match. In general, when you win, your MMR will go up, and when you lose, your MMR will go down. Win/loss is the primary criteria used to update MMR, but individual performance also plays a role, especially when our uncertainty about your MMR is high. It is possible for an individual MMR to increase after a loss or decrease after a win, but in general the winning team’s average MMR will increase and the losing team’s MMR will decrease."
I just played this game, http://dotabuff.com/matches/439389396 and as you can see I played very well. Unfortunately I couldn't get the team to overcome a very good tinker/furion push so we lost. So I lose, big whoop. You win some you lose some. But guess what happens to my matchmaking ranking...well it sure doesn't go up. In fact it goes down -27, more than it has in some other games where I didn't play as well. Listen Valve, don't lie to your devoted audience. If you win you get 25 points if you lose you lose 25 points and then based on your performance (which is calculated badly) you tack on or take off another 1-3 points. Looking at the changes in my ranking has made this very clear (never had a change less than 20 or greater than 30). Yeah the system is flawed but what really bothers me is Valve trying to act like they've created this advanced system that is really quite primitive. Tell us the full truth and I'll take it for what it is. Until then, I'm done being penalized for teammates that drag me down, or moving up when I don't deserve to because I got carried by my teammates.