|
|
On December 05 2013 19:34 hfglgg wrote:Show nested quote +On December 05 2013 19:27 hariooo wrote:On December 05 2013 19:20 hfglgg wrote:
even if they could predict the outcome of a match, it doesnt mean the individual scores are comparable. in an earlier version the two accounts of a friend of mine had a difference of about 500 points with the by valve lower ranked account having the higher number. its save to say that the account with the higher DSR rating, but lower MMR will win more often than the other account with lower DSR but higher MMR because it is underrated.
you can not compare the two ratings (DSR of both accounts) though.
There's really no point talking about earlier models. They were changed because they were bugged, according to the guy at least. And of course you can compare two accounts' DSR's? World of Tanks has a non-ELO rating system that their players use to compare skill level in a team game. DSR is not exactly a unique concept. The creator just has to adapt it to DOTA's unique performance metrics and determine what parameters are significant. my underlying point still stand that its not neccessarily possible to compare two ratings only because they can predict the outcome of a match. so the whole "we can predict xyz (even if true)" doesnt matter anything unless they provide full information about what DSR is.
So i found this post on reddit by the guy developing DSR that gives a little bit more info
http://www.reddit.com/r/DotA2/comments/1s4ezx/mydotaskillcom_and_why_it_should_be_ignored/
Honestly i thought it was a complete joke at the beggining but it might turn out to be legit, i'll see how it goes.
|
On December 05 2013 19:19 njt7 wrote:
Using that logic do you also think anyone clanking down on it has very low DSR or is behind a competing site? I think people should be able to express their opinions on this board without assumptions being thrown in their faces.
Yes. Clearly.
Or maybe its just the way he's just so...intense. And being aggressively insulting to people more or less at random. 
|
On December 05 2013 19:43 hariooo wrote:Show nested quote +On December 05 2013 19:34 hfglgg wrote:On December 05 2013 19:27 hariooo wrote:On December 05 2013 19:20 hfglgg wrote:
even if they could predict the outcome of a match, it doesnt mean the individual scores are comparable. in an earlier version the two accounts of a friend of mine had a difference of about 500 points with the by valve lower ranked account having the higher number. its save to say that the account with the higher DSR rating, but lower MMR will win more often than the other account with lower DSR but higher MMR because it is underrated.
you can not compare the two ratings (DSR of both accounts) though.
There's really no point talking about earlier models. They were changed because they were bugged, according to the guy at least. And of course you can compare two accounts' DSR's? World of Tanks has a non-ELO rating system that their players use to compare skill level in a team game. DSR is not exactly a unique concept. The creator just has to adapt it to DOTA's unique performance metrics and determine what parameters are significant. my underlying point still stand that its not neccessarily possible to compare two ratings only because they can predict the outcome of a match. so the whole "we can predict xyz (even if true)" doesnt matter anything unless they provide full information about what DSR is. What's the evidence for your statement? Can you give me a good example of when you can't compare 2 DSR's if they are valid parameters to predict wins?
i.e. you could use data outside of the only thing that really matters (win or lose) that are affected by the bracket you play in without counting for those. if you, lets say, take GPM as a metric and weight the rating on GPM you will still get a reasonable close prediction because teams with a higher GPM and thus players with a higher GPM win more games. it doesnt account for who you are playing against though, so your might have 700GPM in every game because you only play against comparably bad players due to you playing in a stack with equally bad players all the time.
the combined rating will still be better than mere guessing as good players tend to get higher GPMs in their wins and their losses compared to worse players, but consistent 700GPM doesnt make you burning.
i never trust rating systems that are vague and/or overly complicated. if it takes anything else into account than the outcome of your game and the opponent you are playing against, i always have to ask why they do it complicated when ELO does exactly the same.
ELO works for teamgames as it works for sologames, because the only varying variable (teammates) doesnt effect the outcome in the long run.
so why not just do ELO?
|
On December 05 2013 19:52 SnowfaLL wrote:
lol before my DSR was 600 and 80 percentile.. then after the rework, I am 1496 and 0.1 percentile.. What a joke this thing is.
Either way, I play only support so my KDA and GPM/XPM is much lower than others, therefor any metric calculating those in is false. The only thing that should matter is wins.
Your KDA/GPM/XPM on a hero is now compared to other people's KDA/GPM/XPM with that same hero. Hence the sudden increase in DSR for support players.
It is a nice metric to measure skill relative to others. Certainly much better at measuring an individual's ranking in a 5v5 game than only looking at wins.
Of course other factors will be considered as well. The reddit post some posts above this one has a reply from the creator with more details:
+ Show Spoiler +Hello, Creator here. You are assuming it's impossible to use anything else than Elo based system. I get this criticism all the time - You can't predict players placement in the MM without access to full player data! Of course I can't. I can't if I use Elo. But I am NOT using Elo. That's how it's possible to estabilish your bracket. What I am doing is looking at xx of your last games ( 20 right now, soon will add other options and a chart to track your progress ). Then, I take each of those games, take as much parameters of each one of them and convert each one of those features into hyperdimensional space. I then use special a special kernel function to separate this hyperspace into two parts. Pro games and games of a normal, mortal player. How does it look like? In the basic version: But in many more dimensions. Then, I calculate where each one of your games is placed. The more sure I am your game was pro-level, the more points you will get for it. The more sure the model is that this is some "scrub-level-game" the less points you will get. It's absolutely atrocious having people that didn't spend one hour on this problem counter every counter-argument with "But it's not Elo!" Guess what, Elo sucks. Guess what, with private data, Elo sucks even more. Right now, by calculating average DSR of the two teams in a match, I was able to predict the outcome in the 65% of cases. Supposedly, Valve's ELO matches you each time with around 50% chance of winning. This means one of those things: Non Elo based ranking is better way to matchmake in team games. Elo may be great for individual skill assessment but that doesn't make it the best solution for team-oriented game. Even Gabe Newell said they are NOT satisfied with their system and are looking for something new. Valve IS enforcing the 50% winrate rule. In one way or another, they force a 50% winrate by sometimes making you take really unfair matches. I highly doubt it. Please refrain of this Elo argument. Elo is not the end of the world and it's absolutely NOT the only way of doing things. Mydotaskill.com is still in an early phase. My next step is combining the first iteration of algorithm ( your individual hero performance ) with this latest generation ( that assigns you a Valvesque MM rating ) to create one, robust rating algorithm that accounts both for individual skill, team oriented nature of the game AND fixes the weakness of the Elo solution. Best regards
|
On December 05 2013 19:53 -Celestial- wrote:Show nested quote +On December 05 2013 19:19 njt7 wrote:
Using that logic do you also think anyone clanking down on it has very low DSR or is behind a competing site? I think people should be able to express their opinions on this board without assumptions being thrown in their faces. Yes. Clearly. Or maybe its just the way he's just so...intense. And being aggressively insulting to people more or less at random.
Instead of being a passive aggressive beta like what you've been doing?
On December 05 2013 19:52 SnowfaLL wrote:
lol before my DSR was 600 and 80 percentile.. then after the rework, I am 1496 and 0.1 percentile.. What a joke this thing is.
Either way, I play only support so my KDA and GPM/XPM is much lower than others, therefor any metric calculating those in is false. The only thing that should matter is wins.
Every complaint here is answered quite clearly in the FAQ and other locations. But "what a joke this thing is".
On December 05 2013 19:58 hfglgg wrote:Show nested quote +On December 05 2013 19:43 hariooo wrote:On December 05 2013 19:34 hfglgg wrote:On December 05 2013 19:27 hariooo wrote:On December 05 2013 19:20 hfglgg wrote:
even if they could predict the outcome of a match, it doesnt mean the individual scores are comparable. in an earlier version the two accounts of a friend of mine had a difference of about 500 points with the by valve lower ranked account having the higher number. its save to say that the account with the higher DSR rating, but lower MMR will win more often than the other account with lower DSR but higher MMR because it is underrated.
you can not compare the two ratings (DSR of both accounts) though.
There's really no point talking about earlier models. They were changed because they were bugged, according to the guy at least. And of course you can compare two accounts' DSR's? World of Tanks has a non-ELO rating system that their players use to compare skill level in a team game. DSR is not exactly a unique concept. The creator just has to adapt it to DOTA's unique performance metrics and determine what parameters are significant. my underlying point still stand that its not neccessarily possible to compare two ratings only because they can predict the outcome of a match. so the whole "we can predict xyz (even if true)" doesnt matter anything unless they provide full information about what DSR is. What's the evidence for your statement? Can you give me a good example of when you can't compare 2 DSR's if they are valid parameters to predict wins? i.e. you could use data outside of the only thing that really matters (win or lose) that are affected by the bracket you play in without counting for those. if you, lets say, take GPM as a metric and weight the rating on GPM you will still get a reasonable close prediction because teams with a higher GPM and thus players with a higher GPM win more games. it doesnt account for who you are playing against though, so your might have 700GPM in every game because you only play against comparably bad players due to you playing in a stack with equally bad players all the time. the combined rating will still be better than mere guessing as good players tend to get higher GPMs in their wins and their losses compared to worse players, but consistent 700GPM doesnt make you burning. i never trust rating systems that are vague and/or overly complicated. if it takes anything else into account than the outcome of your game and the opponent you are playing against, i always have to ask why they do it complicated when ELO does exactly the same. ELO works for teamgames as it works for sologames, because the only varying variable (teammates) doesnt effect the outcome in the long run. so why not just do ELO?
One of the first things you learn in statistics about modifying variables is that all other variables have to be held constant. So the affect of winrate on ELO assumes everything else (like the teammate variable) is constant. This is clearly not the case since teams change. Thus you need a more complicated model to deal with issues like that.
|
Dude. Chill. Seriously.
The last two iterations of DSR were bugged, yeah? They've been improved on since, yeah? I'd say it's statistically very probable that this iteration is ALSO imperfect and that the work these dudes have put in so far indicates they're likely to improve it even further. There's no need to aggressively defend DSR. Stop it.
|
I dont think KDA is legit way to measure players' skill
|
On December 05 2013 20:13 hariooo wrote:Show nested quote +On December 05 2013 19:53 -Celestial- wrote:On December 05 2013 19:19 njt7 wrote:
Using that logic do you also think anyone clanking down on it has very low DSR or is behind a competing site? I think people should be able to express their opinions on this board without assumptions being thrown in their faces. Yes. Clearly. Or maybe its just the way he's just so...intense. And being aggressively insulting to people more or less at random. Instead of being a passive aggressive beta like what you've been doing? Show nested quote +On December 05 2013 19:52 SnowfaLL wrote:
lol before my DSR was 600 and 80 percentile.. then after the rework, I am 1496 and 0.1 percentile.. What a joke this thing is.
Either way, I play only support so my KDA and GPM/XPM is much lower than others, therefor any metric calculating those in is false. The only thing that should matter is wins. Every complaint here is answered quite clearly in the FAQ and other locations. But "what a joke this thing is". Show nested quote +On December 05 2013 19:58 hfglgg wrote:On December 05 2013 19:43 hariooo wrote:On December 05 2013 19:34 hfglgg wrote:On December 05 2013 19:27 hariooo wrote:On December 05 2013 19:20 hfglgg wrote:
even if they could predict the outcome of a match, it doesnt mean the individual scores are comparable. in an earlier version the two accounts of a friend of mine had a difference of about 500 points with the by valve lower ranked account having the higher number. its save to say that the account with the higher DSR rating, but lower MMR will win more often than the other account with lower DSR but higher MMR because it is underrated.
you can not compare the two ratings (DSR of both accounts) though.
There's really no point talking about earlier models. They were changed because they were bugged, according to the guy at least. And of course you can compare two accounts' DSR's? World of Tanks has a non-ELO rating system that their players use to compare skill level in a team game. DSR is not exactly a unique concept. The creator just has to adapt it to DOTA's unique performance metrics and determine what parameters are significant. my underlying point still stand that its not neccessarily possible to compare two ratings only because they can predict the outcome of a match. so the whole "we can predict xyz (even if true)" doesnt matter anything unless they provide full information about what DSR is. What's the evidence for your statement? Can you give me a good example of when you can't compare 2 DSR's if they are valid parameters to predict wins? i.e. you could use data outside of the only thing that really matters (win or lose) that are affected by the bracket you play in without counting for those. if you, lets say, take GPM as a metric and weight the rating on GPM you will still get a reasonable close prediction because teams with a higher GPM and thus players with a higher GPM win more games. it doesnt account for who you are playing against though, so your might have 700GPM in every game because you only play against comparably bad players due to you playing in a stack with equally bad players all the time. the combined rating will still be better than mere guessing as good players tend to get higher GPMs in their wins and their losses compared to worse players, but consistent 700GPM doesnt make you burning. i never trust rating systems that are vague and/or overly complicated. if it takes anything else into account than the outcome of your game and the opponent you are playing against, i always have to ask why they do it complicated when ELO does exactly the same. ELO works for teamgames as it works for sologames, because the only varying variable (teammates) doesnt effect the outcome in the long run. so why not just do ELO? One of the first things you learn in statistics about modifying variables is that all other variables have to be held constant. So the affect of winrate on ELO assumes everything else (like the teammate variable) is constant. This is clearly not the case since teams change. Thus you need a more complicated model to deal with issues like that.
the teammate variable is constant. the pool yours and your opponents players are choosen from is the same and the players are divided more or less at random. thus you get the same players on your team and on the opposing team eventually in every possible configuration an indifinite amount of times, leaving only your personal performance to determine your rating.
|
|
|
He doesn't need to make the algorithm public.
All that will do is cause a million loud, uninformed people to make silly counterarguments. He's obviously using some fairly arcane statistical methods that the average dota player won't understand. Of course, doing something complicated doesn't automatically make him correct, but it does mean that 99.9% of his audience isn't qualified to have an opinion based on the methodology, so releasing it is irrelevant for most of us.
What we are qualified to have an opinion on is whether it makes sense in terms of results. If he really wants to prove himself, he should walk through some cases where he demonstrates his predictions and outcomes, and make sure the cases include some smurfs and high-disparity stacks. That's the kind of information that a layman audience can actually process.
Personally I think this whole thing is a scam to get Yango to blow his 20k by banning everyone for surreptitiously discussing ranked ladders.
|
I love one of the advice to improve your DSR: "Be a team player, as a carry - help your teammates with ganks and don't die.
" Noob carry no gank"
|
On December 05 2013 19:01 hariooo wrote:Show nested quote +On December 05 2013 18:45 BurningSera wrote:On December 05 2013 17:58 hariooo wrote:On December 05 2013 17:42 BurningSera wrote:On December 05 2013 17:08 hariooo wrote:On December 05 2013 15:44 njt7 wrote:
The only way to comparatively use DSR effectively is against yourself for later references. Comparing against others wont give you anything good. And the site is still in massive beta from the looks of it so your DSR number might change just for the worst just from patching.
The developer seems decent so in the future it might be useful but all of these different rating numbers is usually worthless for self improvement its dangerous to get caught in pointless ego boosting. If people actually want to improve their play they should watch their own replays and analyze as well as watching specific heroes/roles played by the best(pros) and try to emulate. Most important is probably to try to find people to play with/ against who is already better then you.
Not sure why I went on this rant but Im getting tired of this DSR/Dotabuff mania going on. Of course you can compare it to other people. It's also the whole basis of why DSR can predict the winner of a game. On December 05 2013 15:48 Staboteur wrote:On December 05 2013 15:21 FakeDeath wrote:
So I just tried out this mydotaskill.com thing everyone is talking about.
My DSR is 954 Emerald whatever that means.
Can this be taken as a legit base line of your skill level?
Not like a true indicator but more of where u stand in the dota ladder?
Though I wonder how they actually calculated your DSR.
What is the difference between this and Valve internal rating? DSR's gone through 3 or so iterations where they adjust the algorithms to produce new and exciting numbers that may or may not mean a damn thing to anyone at all. They're not a good indicator of skill. By the site's only admission, it is only able to predict the winner of any given game correctly 65% of the time. Seeing as that's only 15% better than a blind guess over many iterations, I'd say it is still pretty damn inaccurate. People who don't have a background in statistics need to stop saying things like this. 65% prediction accuracy over a large sample is RIDICULOUS and if that 65% is proven to be true it is clearly evidence of DSR's accuracy. It would be a statistical fact except people who don't study stats can't be convinced because they can't understand the proof. I dont think any of that matters when most of the player profile is set to be hidden by default. Meanwhile, Player number 34176943 DSR is: 1275 This player has a Pro DSR score (top 0,5%). >_> and my smurf now is 460. I am entertained enough by now, that will be the last time i visit the site. and i fcking respect the people behind dotabuff now. http://www.reddit.com/r/DotA2/comments/1s4ezx/mydotaskillcom_and_why_it_should_be_ignored/cdtubbcHidden profiles shouldn't matter. Yes it does very much so matters, unless some people just wanted a number to say something meaningless about their skill/ranking then it shouldn't matter to them i guess (or anything matters to them in that case, hence the ridiculous ranking from the website). You can talk about fancy stats techniques etc but in the end of day I hope people know that the site is a joke lol (a bigger joke than DBR if you must, and i actually respect the stats guys behind dotabuff now because they actually have the self-respect and morale as some statisticians. say 1000 out of 100000 players are actual top class player, ill give you 800 of them have the profile opened for public because high level players surely care about their ranking (obviously i am being very generous here, go search dendi pub games you will almost always find 1-5 players that are anonymous), so in this case there are 200 top class players that will never be included in the stats results, ie the ranking of these 800 players will forever be not accurate because we know there are differences between top top top top class players and top class players. and then lets say 40000 out of the remaining 99000 players have made their profile public, that's a very generous number because that would mean almost 40% of the players are showing their profile. I think I'll stop here now because if you still think that DSR means anything then there is nothing more i say can persuade you to think otherwise lol. p/s: i am guessing that DSR is using somesort of 'curve fitting' model to 'predict' people rankings but in this case of dota2 player pool, it wont be a meaningful curve. Holy fuck. You're trying to argue against something that already happens in real life. Let me see if I can break this down really simply for you with a braindead example. When your country (I'm assuming the UK does these) conducts a census of the population, they do not collect data from every single person living in the UK. They collect an INCOMPLETE data set. Yet they can still draw conclusions from that data set like who the expect certain demographics to vote for, if there is mass migration from rural to urban districts, etc. Here's another example because of how adamantly you want to argue on a topic with which you clearly have no academic or professional experience. When Nate Silver perfectly predicted the 9 swing states for the US 2012 election, I PROMISE YOU he did not collect data from every single voter in those states. Drawing accurate conclusions from incomplete data is done ubiquitously all over the world in many different fields. DOTA2 is just one of them.
You still dont get it do you. I'll put it in another angle to read my post:
'DSR is for some people who just wanted a number to say something about their skill/ranking'.
Of course it depends on how well one is convinced from that website but in reality that 'something' can be pretty meaningless. How meaningless it can be? nobody knows. If the creator of the website said the accuracy is X% then he is pulling that out from his ass, he CANNOT know that X%.
I am not even talking about the absolute accuracy here, i am talking about a convincing ranking. The fact that the website is telling people to stop stacking with noob friends in order to get a higher ranking means that they cant be sure of your skill level if you stacked. Valve who has all the real stats couldn't figure out a good MM to work (they will need to make a good ranking system behind to do so) and some magical website somehow capable to achieve that?
Funny because i am doing some population-based cohort study of some patients of a city so clearly I don't know anything about statistics.
|
Imho they should just shut it down.
Not that i dislike Rankings, i dislike everyone talking non stop about it (see this very tread)... it's just boring and stupid.
|
On December 05 2013 20:13 hariooo wrote:Show nested quote +On December 05 2013 19:53 -Celestial- wrote:On December 05 2013 19:19 njt7 wrote:
Using that logic do you also think anyone clanking down on it has very low DSR or is behind a competing site? I think people should be able to express their opinions on this board without assumptions being thrown in their faces. Yes. Clearly. Or maybe its just the way he's just so...intense. And being aggressively insulting to people more or less at random. Instead of being a passive aggressive beta like what you've been doing?
First of all, calm down.
Second, you want me to be more direct? Fine by me. Remember, you're the one who decided to jump on me with your inane and bloody obvious comment that "what the numbers mean has changed!" Then you decided to follow it up with a flippant comment about "oh you've got to have soloQ games that aren't in LP" which frankly I find hillariously presumptive and, again, bloody obvious.
Not to mention your high-handed and insulting responses with regard to anyone in here who has actually attempted to ask you a question or, god forbid, actually make a point against it. You're arrogant, aggressive and even being nasty to those genuinely curious.
No, not everyone on here has statistical experience to understand arguments in favour of the system (I do as a matter of fact, I'm a research scientist, but that's besides the point); but I'm willing to bet a good proportion of people on here are smart, educated people with the CAPABILITY to understand. However the way you're treating people is convincing nobody of anything. Sort out your attitude and people may start to actually listen.
|
On December 05 2013 21:32 Jetaap wrote:
I love one of the advice to improve your DSR: "Be a team player, as a carry - help your teammates with ganks and don't die.
" Noob carry no gank"
Seriously. Any rating not based purely on if you win is failure. Yes, teammates make it harder to judge whether someone is winning because of themselves or others, but that's the whole point of large sample sizes and having individual ratings that contribute to determining a team's chance of winning.
Changing the tone a little, can couriers use Linken's? Like, could you have a flock of couriers casting Linken's on people when needed? Watching Hao's 6 slot AM where he has no idea what to do with money from last night's D2L game puts me in the mood for ultra-hyper-super late game theorycrafting.
|
On December 05 2013 21:38 Velr wrote:
Imho they should just shut it down.
Not that i dislike Rankings, i dislike everyone talking non stop about it (see this very tread)... it's just boring and stupid.
its a bait for yango to waste his 20k post
|
H4nn1 @KaiH4nn1
Follow
Not to confuse ppl here but it was not on our end that we didn't get the visas done in time for china, I would have loved to go )
I'm dumb and can't do the cool twitter embed thing. Anyways was this announced anywhere? I assume this is for g-league? Is anyone replacing them g-league is like a week from now no?
|
United Kingdom38257 Posts
Both fnatic and lgd.int are out due to visa stuff.
They'll just downsize the groups to accommodate.
|
On December 05 2013 21:48 FuzzyJAM wrote:
Changing the tone a little, can couriers use Linken's? Like, could you have a flock of couriers casting Linken's on people when needed? Watching Hao's 6 slot AM where he has no idea what to do with money from last night's D2L game puts me in the mood for ultra-hyper-super late game theorycrafting.
Hmm...thats got me thinking...does BKB maintain when you drop the item? I mean if you activate it and then drop your BKB.
Because in THEORY with something like Antimage you could activate BKB and then switch out for another, better stats, item really quickly on a courier carrying your stash before blinking in to fight.
|
On December 05 2013 21:25 Belisarius wrote:
He doesn't need to make the algorithm public.
All that will do is cause a million loud, uninformed people to make silly counterarguments. He's obviously using some fairly arcane statistical methods that the average dota player won't understand. Of course, doing something complicated doesn't automatically make him correct, but it does mean that 99.9% of his audience isn't qualified to have an opinion based on the methodology, so releasing it is irrelevant for most of us.
What we are qualified to have an opinion on is whether it makes sense in terms of results. If he really wants to prove himself, he should walk through some cases where he demonstrates his predictions and outcomes, and make sure the cases include some smurfs and high-disparity stacks. That's the kind of information that a layman audience can actually process.
Personally I think this whole thing is a scam to get Yango to blow his 20k by banning everyone for surreptitiously discussing ranked ladders.
Yango is cheating, he has created a new account despite that it's forbidden. It's called notyango.
|
|
|
|
|
|
EPT
