^ I'm having the same problem as that guy. It's been a couple of days now.
"Please wait a while after submitting your request. There is a waiting queue for the requests.
Player number 93803801 skill rating is still being calculated. Currently, your position in the queue is .
Each hour, 6000 players profiles are processed. Once you reach position one, you will be able to see your score here."
Lol even though my computer has crashed in my last 20 games(lowering the score by a lot), I really dont think I have such a bad score. Not really sure what he is basing it on, he said something about playing selfish(I might everynow and again), but I always will try to gank and win my lane.
Stay tuned. I will update it soon. There are over 100,000 player profiles that need to be updated :)
It's highly possible you will be able to see your past DSR on a chart, game by game.
Also, I plan to introduce individual hero metrics - how good you are with any particular hero. Then I will create a ranking of best players for each of the heroes.
Well winrate do not matter. I created this account for fun to see what I would get on a new account. I haven't won very much, but I still got around the same amount OF DSR as I got on my main account.
I think I had 11/9 or 10/10 when I was done with the 20 games and I received a score on 14xx. (Not stacking - all 20 games solo)
This DSR is totally useless. I play with friends that goes from 200 matches to 1500. We use to play as a group, so usually 2 to 5 peoples.
We have 2 newbies playing with us. We can totally wreck them in pure skill, reaction, vision, teamplay, anticipation, etc. Yet this tool give them twice our scores.
So, to me this "tool" is "useless" because it doesn't reveal/reflect the true skill/templay of what is really going on in a dota 2 match.
@Mastah, did you consider that your ratings were skewed because you stacked with players of a different skill level?
Do you understand that your actual skill does not matter, what matters is what you put into a game.
ie. Are you going to call the algorithm used by whoscored crap, because it gave Vidal and Neymar higher ratings in individual matches while Ronaldo is clearly a far superior player? Just think man, fuck.
If a 1700 player created a new account, and made no effort at all, is the system shit for giving them a lower rating? Nope, its doing its job. Now if you used your fucking eyes you would realise that this system only works for solo queue players, and is still a dynamic system which can be influenced by queueing at strange times, and can be especially volatile in higher level matchmaking because there are so few people at this level that would be in the queue at the same time, and probably has the largest range of ratings in each match.
Terrible: A rating system such as this is supposed to calculate a player's ability. As you can see from my previous lengthy post (and many others here) it is shockingly bad at doing so, so yes - the algorithm is flawed. On the DSR website, it suggests a few ideas to help you improve your performance, such as helping with ganking and not being too greedy, but the trouble with this is that Dota is incredibly situational and what the team needs will vary depending on the strategies employed. It would seem that if the system disagrees with these strategies it ranks a player's ability negatively. A player's skill level is not measured by the number of wards placed when playing a support role. Personally, I believe the only real, accurate way for a third party software to conclude a player's rating would be to build a database of players and judge it on a who-beat-who basis.
@Trillion, it clearly says that it does not work for players that don't solo queue. And of course no system is going to be perfect, but this system is clearly doing something right when people that solo queue onto first page games often are getting high ratings, and people with incredible stats (KDA/win rate) that play in lower level matches get a lower rating.
ELO is a shitty system, and easily exploitable. It is good for 1v1 games, like chess, not for 5v5 matches of random players (wouldn't be too bad for team MM). Its better than nothing, but a dynamic system similar to DSR is a far better system, making it precise is not possible because a player's performance will vary from game to game, especially with such a diverse hero pool. So while someone like Admiral Bulldog might be a ridiculous player when he uses Furion/LD/Clock or whoever, with another hero many other players will be better than him, what you are saying is that this should not be the case
A good rating system does not bias your performance as you for some reason, think that it should, a good rating system judges you based on your performance (individual + win/loss) weighted against the skill level of the game.
This thread is filled with two kinds of idiots;
1. People that are shit, and are surprised by their low rating. One of the guy's on this page (2nd post) primarily plays in the normal bracket (~55% of his matches), quite a lot of that would be because he queues with lower ranked players, regardless, he is surprised by his 200 rating.
2. People that don't read, like yourself. A system is not unsuccessful if it doesn't do something that its not supposed to do. I can't drive my toothbrush to work, is my toothbrush unsuccessful? It clearly says that it only works for solo queue players, and if you want an accurate rating, then play solo queue for 20 matches.
Also most of your arguments are situational, and this is important for you to understand that while there are situations which may seem "negative" as far as the data that is used in his system is concerned, there are two things you need to consider. First off, that the win/loss is more important than anything else, if you are sacrifcing yourself or baiting, and your team still wins...then what is the problem? The other point is that its based on 20 matches, maybe a little more in the future, and on average you won't be doing such things anyway.
With that in mind, take that massive stick out of your ass and either ignore the website or play solo queue for 20 matches to see where you really stand. From what I have seen, other than a handful or smurfs getting very high ratings (some are deserved because they are actually playing first game matches and doing well), most players are getting proper ratings from this site. You can look at the exploiters in LPQ getting ridiculous scores and only having 1000 or less rating (not 200 because they encounter another 5 virigin russian stackers in LPQ), and then there are players with 50% win rate, low (2-3) KDAs that play page 1-4 games often that have a very high rating.
I am very curious about this algorithm. The fact that DSR ratings can predict wins 65% caught my attention. If that is really true, I am amazed and very very curious to learn how it works. I want to offer support in improving the prediction percentage. I am a Ph.D student in an applied math sort of an area. Coding, Probability theory, algorithms are routine for me. So please be as technical as you want while explaining the idea.
No matter what people in this thread say, i believe the true power of a theory lies in its predictive power. If DSR can predict win rates with more than 50% chance apriori, then it's already good. We have to learn to make it better.
With the blog update from valve, we can ask new questions.
1) Are you using variables like 'time integral of the gold earned from the last 0 crossover' to estimate the closeness of the match? Or does your method only use the visible parameters of the Dotabuff site?
2) I can see why stacking can completely screw estimating individual skills. So do you have a handle on checking whether somebody stacks or not? Valve's idea of having two separate MMR at least attempts to separate the problem.
3) Estimating the joint MMR distribution of 5 players by knowing *only* their individual MMR's would need some additional assumptions on the data. It seems true that individual MMR behaves like a normal distribution (from Valve's blog update). But does the joint distribution look like multivariable normal?
I have so many questions on this and want to help in ways to improve the estimates. If you are generous enough, please share your thoughts on my doubts. I am a new player who sucks at Dota, but I like playing the game. I like maths and algorithms even more and luckily, I am not bad at that :P
The first iteration that I've released was trying to assign DSR based on individual hero performance - How are your Kills, Assists, Deaths, XPMs, GPMs, Winrate in relation to other players of the same hero. Additionally, it looked at your opponents winrates to try to guess where you play ( this proved to be extremely noisy ).
This approach resulted in wrong results for smurfs, heavy stackers and people who use their accounts to play with their worse-skilled friends.
After 2 iterations of this algorithm I've come to conclusion it's pointless to analyze those stats in vacuum - you can't really know how to determine where somebody plays ( and to achieve some kind of parity with Valve's internal matchmaking score, you really have to guess where they play ).
So I've discarded all hero metrics and focused instead of looking at any players last 20 games and trying to guess on what level were they played. If you look at two matches - one of a beginner and one of a pro player, there are some really visible(and also some subtle) differences. If you average the stats of last 20 games and then run SVM classifier on them, you will be able to distinguish pro from a mere mortal in almost 100%.
Because I use ranking SVM, I get not only the binary decision ( is this a pro player or a normal player )? But also I get the score - how much is this player a pro player / how much is this player a normal player? (By measuring the distance from hyperplane)
Unfortunately, right now, this approach is heavily skewed by stacking. If you stack with a worse friend, your DSR rating will plummet. If you stack with your good friends, your DSR will skyrocket - all because "pro-likeness" of each of those games changes dramatically - since you are being taken out of your normal player base into (presumably) 8-9 players from a wildly different bracket. Those players games result in certain statistics that can be used to approximate the level of their games.
Right now, DSR works best if you solo queue - because there is no one who would drag you into other brackets.
I plan to combine the first iteration ( hero performance metrics ) with the second iteration ( avg bracket estimator ) to create the third generation of DSR algorithm that would account both for how you fare with each hero and where you play ( this should, at least, slightly alleviate problems with both approaches ).
I welcome anybody who is interested in this subject to talk to me. I will be happy to brainstorm, find new ideas and test them. More so math/stats PhD's!
I've started the first refresh 2 hours ago ( I had to squash some bugs and it still hungs up, I will try to move the whole thing to nginx and it should run faster ).
So far we sit at 5000 scores refreshed out of 110.000 - it goes rather slowly. I will try to improve the speed so the refreshes will come daily.
There is still SO much to add to the DSR formula - if I were to estimate, I've added only around 20% of variables that could be used to estimate your bracket. So what people really see right now is still an extremely early version. Imagine how much more accurate it should be if you add the rest of the variables.
If you want some help with the website design, maybe I can offer some help. I have no clue about programming, but I know something about Photoshop and stuff.
a) It's a fun project and I enjoy working on it.
b) Valve's rating will not be public. You will be able to see only your own MMR and those of your friends.
c) Valve's relying on Elo and I want to eventually develop something that works differently.
d) There's plenty of things I plan to add besides DSR rating.
What you say is completly false and seems to show that you do not understand how the game and how matchmaking works.
Recently Valve has made a blog post about how matchmaking works, you should definitely read it before saying false statements like you did.
Valve said that they try to make every game even via matchmaking it mean that if they manage to do it for every game each team has 50% of win in every game. So just by fliping a coin you should be able to predict 50% of the winning team (which you could do even without their matchmaking system).
But it's not their only priority when doing matchmaking they also have different criteria like making the waiting time shorts, puting players from same country together....
Basically it means the 50% value they want achieve is only theoretic due to these others criteria and due to the fact that their MMR models can't predict everything in a game. For example their system can't take account if a player is improving his MMR and will likely have a way higher MMR in the next 50 games.
So let's assume for every particular game the real match up odds is more likely to be often in the 30%-70% range than to be 50% acurately, meaning that if the same both team play 100 times one will win beetween 30%-70% of the times instead of this 50% ideal and theoretic value; which is actually quite a natural assumptions.
And now knowing than someone has plenty of data to analyse the two teams and can actually see if the match up is indeed exactly even or a bit one sided, predict 65% isn't that impressive at all; 85-90% would be impressive.
I don't think you understand basic statistics. "Each team has a 50% chance to win." It's a value they can only garantee on average, they can't garantee to put the odds for a particular game at 50%, but you try to predict wich team will win for a particular game.
Not only that but it's only one of their criteria.
I played solo queue for my last 45 games, my rating suddenly dropped from 9xx to 4xx. I won some, I lost some, my performances were normal. I'd like to know finally where I am standing, but this rating won't tell me that. I sometimes checked people in my games and I had 1100-er guys and 200-er guys
I hope Valve doesn't just give us a rating but makes it visible, where you are actually standing. I don't even care a lot about other people, I want to know my rating and that mine is right. This thing tells me apparently not a lot.
Also that: http://mydotaskill.com/index.php?id=18035715 Owner is apparently god of Dota 2: 3594 points...
Maybe I need to wait until all scores are updated? oO Or it's just very bugged? oO
Ok so it definitely doesn't work, I did not stack at all. Ok at least I know now.
Edit: Yes if I stacked I might be lower now I understood that. But I didn't do so or do you think I played every game in another stack? I even chose "Do not match with parties when alone". And you updated your comment, suddenly I did not stack in your data. Well that's true but how it comes you've seen it like this before. And your own score dropped in 10 minutes from 3500 to 650, wow.
Dsr is accurate if you are going to play last 20 games solo queue. When I played 20 games solo queue was 1550 Dsr,when I played last 10 with some friends I droped fast to 750 = ) .
@Captain
Well, I've checked and there's plenty of 48% winrate and 1.5 - 2 kda people in the matches. Also, the only team matchmaking rating I could find was 2500 ( which is quite low ).
But maybe you are an outlier and I'd have to investigate further.
I do not want to say I was at 900 or 400 or 1500 or 100, like I said I dont know where I am.
But I know: I did not stack for 45 games, I only played with equal or worse players before, I never played in team-matchmaking, I have no team (or do you mean when you stack with 1 or 2 it's team-matchmaking already?). So either matchmaking somehow throws me suddenly with much worse players in a game when I play at normal times in normal solo queue All-Pick or something else is wrong.
Lost more games than I won in the last week for sure, but my DSR went up 108 point from 771 to 879. I have been queuing up with better players but have also won some of those matches. i still play solo more than stacked however.
Emerald now. I thought that was top 3% or something... not sure. I remember it was platinum, emerald, pro, mythril
^ I'm having the same problem as that guy. It's been a couple of days now.
"Please wait a while after submitting your request. There is a waiting queue for the requests.
Player number 93803801 skill rating is still being calculated. Currently, your position in the queue is .
Each hour, 6000 players profiles are processed. Once you reach position one, you will be able to see your score here."
My ID# 93803801
wtf, i have 200.
Come on, i`m not that bad at this game
Does the creator have a response/explanation for these discrepancies?
Yes, if you have played with lower skilled people or if you have played selfishly then you will receive a lower score.
Read my previous post. Thanks.
Lol even though my computer has crashed in my last 20 games(lowering the score by a lot), I really dont think I have such a bad score. Not really sure what he is basing it on, he said something about playing selfish(I might everynow and again), but I always will try to gank and win my lane.
Too sad that it doesn't update the score :(
Stay tuned. I will update it soon. There are over 100,000 player profiles that need to be updated :)
It's highly possible you will be able to see your past DSR on a chart, game by game.
Also, I plan to introduce individual hero metrics - how good you are with any particular hero. Then I will create a ranking of best players for each of the heroes.
Best regards
Thank you, Fear.
Player number 76121868 DSR is:
1621
This player has a Mithril DSR score .
is this good? theres no information about ranking and stuff
@beld
Sure, I have a score of 300>.>
Well winrate do not matter. I created this account for fun to see what I would get on a new account. I haven't won very much, but I still got around the same amount OF DSR as I got on my main account.
I think I had 11/9 or 10/10 when I was done with the 20 games and I received a score on 14xx. (Not stacking - all 20 games solo)
This DSR is totally useless. I play with friends that goes from 200 matches to 1500. We use to play as a group, so usually 2 to 5 peoples.
We have 2 newbies playing with us. We can totally wreck them in pure skill, reaction, vision, teamplay, anticipation, etc. Yet this tool give them twice our scores.
So, to me this "tool" is "useless" because it doesn't reveal/reflect the true skill/templay of what is really going on in a dota 2 match.
http://mydotaskill.com/index.php?id=57068090 how come im not mythril but my DSR is above 750? wtf?
@Mastah, did you consider that your ratings were skewed because you stacked with players of a different skill level?
Do you understand that your actual skill does not matter, what matters is what you put into a game.
ie. Are you going to call the algorithm used by whoscored crap, because it gave Vidal and Neymar higher ratings in individual matches while Ronaldo is clearly a far superior player? Just think man, fuck.
If a 1700 player created a new account, and made no effort at all, is the system shit for giving them a lower rating? Nope, its doing its job. Now if you used your fucking eyes you would realise that this system only works for solo queue players, and is still a dynamic system which can be influenced by queueing at strange times, and can be especially volatile in higher level matchmaking because there are so few people at this level that would be in the queue at the same time, and probably has the largest range of ratings in each match.
does this site ever update?
Concede, never, my score was the same week ago.
Terrible: A rating system such as this is supposed to calculate a player's ability. As you can see from my previous lengthy post (and many others here) it is shockingly bad at doing so, so yes - the algorithm is flawed. On the DSR website, it suggests a few ideas to help you improve your performance, such as helping with ganking and not being too greedy, but the trouble with this is that Dota is incredibly situational and what the team needs will vary depending on the strategies employed. It would seem that if the system disagrees with these strategies it ranks a player's ability negatively. A player's skill level is not measured by the number of wards placed when playing a support role. Personally, I believe the only real, accurate way for a third party software to conclude a player's rating would be to build a database of players and judge it on a who-beat-who basis.
probably for the better, i've been losing too much.
@Fearohnum // MyDotaSkill.com
I like the hero metrics idea.
@Trillion, it clearly says that it does not work for players that don't solo queue. And of course no system is going to be perfect, but this system is clearly doing something right when people that solo queue onto first page games often are getting high ratings, and people with incredible stats (KDA/win rate) that play in lower level matches get a lower rating.
ELO is a shitty system, and easily exploitable. It is good for 1v1 games, like chess, not for 5v5 matches of random players (wouldn't be too bad for team MM). Its better than nothing, but a dynamic system similar to DSR is a far better system, making it precise is not possible because a player's performance will vary from game to game, especially with such a diverse hero pool. So while someone like Admiral Bulldog might be a ridiculous player when he uses Furion/LD/Clock or whoever, with another hero many other players will be better than him, what you are saying is that this should not be the case
A good rating system does not bias your performance as you for some reason, think that it should, a good rating system judges you based on your performance (individual + win/loss) weighted against the skill level of the game.
This thread is filled with two kinds of idiots;
1. People that are shit, and are surprised by their low rating. One of the guy's on this page (2nd post) primarily plays in the normal bracket (~55% of his matches), quite a lot of that would be because he queues with lower ranked players, regardless, he is surprised by his 200 rating.
2. People that don't read, like yourself. A system is not unsuccessful if it doesn't do something that its not supposed to do. I can't drive my toothbrush to work, is my toothbrush unsuccessful? It clearly says that it only works for solo queue players, and if you want an accurate rating, then play solo queue for 20 matches.
Also most of your arguments are situational, and this is important for you to understand that while there are situations which may seem "negative" as far as the data that is used in his system is concerned, there are two things you need to consider. First off, that the win/loss is more important than anything else, if you are sacrifcing yourself or baiting, and your team still wins...then what is the problem? The other point is that its based on 20 matches, maybe a little more in the future, and on average you won't be doing such things anyway.
With that in mind, take that massive stick out of your ass and either ignore the website or play solo queue for 20 matches to see where you really stand. From what I have seen, other than a handful or smurfs getting very high ratings (some are deserved because they are actually playing first game matches and doing well), most players are getting proper ratings from this site. You can look at the exploiters in LPQ getting ridiculous scores and only having 1000 or less rating (not 200 because they encounter another 5 virigin russian stackers in LPQ), and then there are players with 50% win rate, low (2-3) KDAs that play page 1-4 games often that have a very high rating.
Hello Dota Skill,
I am very curious about this algorithm. The fact that DSR ratings can predict wins 65% caught my attention. If that is really true, I am amazed and very very curious to learn how it works. I want to offer support in improving the prediction percentage. I am a Ph.D student in an applied math sort of an area. Coding, Probability theory, algorithms are routine for me. So please be as technical as you want while explaining the idea.
No matter what people in this thread say, i believe the true power of a theory lies in its predictive power. If DSR can predict win rates with more than 50% chance apriori, then it's already good. We have to learn to make it better.
With the blog update from valve, we can ask new questions.
1) Are you using variables like 'time integral of the gold earned from the last 0 crossover' to estimate the closeness of the match? Or does your method only use the visible parameters of the Dotabuff site?
2) I can see why stacking can completely screw estimating individual skills. So do you have a handle on checking whether somebody stacks or not? Valve's idea of having two separate MMR at least attempts to separate the problem.
3) Estimating the joint MMR distribution of 5 players by knowing *only* their individual MMR's would need some additional assumptions on the data. It seems true that individual MMR behaves like a normal distribution (from Valve's blog update). But does the joint distribution look like multivariable normal?
I have so many questions on this and want to help in ways to improve the estimates. If you are generous enough, please share your thoughts on my doubts. I am a new player who sucks at Dota, but I like playing the game. I like maths and algorithms even more and luckily, I am not bad at that :P
Thanks for your time.
Hello Isomorphism,
The first iteration that I've released was trying to assign DSR based on individual hero performance - How are your Kills, Assists, Deaths, XPMs, GPMs, Winrate in relation to other players of the same hero. Additionally, it looked at your opponents winrates to try to guess where you play ( this proved to be extremely noisy ).
This approach resulted in wrong results for smurfs, heavy stackers and people who use their accounts to play with their worse-skilled friends.
After 2 iterations of this algorithm I've come to conclusion it's pointless to analyze those stats in vacuum - you can't really know how to determine where somebody plays ( and to achieve some kind of parity with Valve's internal matchmaking score, you really have to guess where they play ).
So I've discarded all hero metrics and focused instead of looking at any players last 20 games and trying to guess on what level were they played. If you look at two matches - one of a beginner and one of a pro player, there are some really visible(and also some subtle) differences. If you average the stats of last 20 games and then run SVM classifier on them, you will be able to distinguish pro from a mere mortal in almost 100%.
Because I use ranking SVM, I get not only the binary decision ( is this a pro player or a normal player )? But also I get the score - how much is this player a pro player / how much is this player a normal player? (By measuring the distance from hyperplane)
Unfortunately, right now, this approach is heavily skewed by stacking. If you stack with a worse friend, your DSR rating will plummet. If you stack with your good friends, your DSR will skyrocket - all because "pro-likeness" of each of those games changes dramatically - since you are being taken out of your normal player base into (presumably) 8-9 players from a wildly different bracket. Those players games result in certain statistics that can be used to approximate the level of their games.
Right now, DSR works best if you solo queue - because there is no one who would drag you into other brackets.
I plan to combine the first iteration ( hero performance metrics ) with the second iteration ( avg bracket estimator ) to create the third generation of DSR algorithm that would account both for how you fare with each hero and where you play ( this should, at least, slightly alleviate problems with both approaches ).
I welcome anybody who is interested in this subject to talk to me. I will be happy to brainstorm, find new ideas and test them. More so math/stats PhD's!
If you are interested, addd me on steam. :)
Best regards
Fearohnum,
Sorry if I echo this question, but how often does the site update? I've been stuck on 811 for like 20+ games already.
I've started the first refresh 2 hours ago ( I had to squash some bugs and it still hungs up, I will try to move the whole thing to nginx and it should run faster ).
So far we sit at 5000 scores refreshed out of 110.000 - it goes rather slowly. I will try to improve the speed so the refreshes will come daily.
Thanks, mate. Also a big thank you for taking all the time to make it work although all the negative feedback :)
Thanks :) People like you help me going.
There is still SO much to add to the DSR formula - if I were to estimate, I've added only around 20% of variables that could be used to estimate your bracket. So what people really see right now is still an extremely early version. Imagine how much more accurate it should be if you add the rest of the variables.
Best regards
If you want some help with the website design, maybe I can offer some help. I have no clue about programming, but I know something about Photoshop and stuff.
but why are u keeping updating ? new mm is comming, whats the point ?
Because:
a) It's a fun project and I enjoy working on it.
b) Valve's rating will not be public. You will be able to see only your own MMR and those of your friends.
c) Valve's relying on Elo and I want to eventually develop something that works differently.
d) There's plenty of things I plan to add besides DSR rating.
@ Isomorphism
"The fact that DSR ratings can predict wins 65% caught my attention."
You should not be that much impressed he predict only 15% better that if you flip a coin.
^ Anyone who says something like this doesn't really understand how matchmaking is supposedly working.
zzzfffffffff
@ Fearohnum // MyDotaSkill.com
What you say is completly false and seems to show that you do not understand how the game and how matchmaking works.
Recently Valve has made a blog post about how matchmaking works, you should definitely read it before saying false statements like you did.
Valve said that they try to make every game even via matchmaking it mean that if they manage to do it for every game each team has 50% of win in every game. So just by fliping a coin you should be able to predict 50% of the winning team (which you could do even without their matchmaking system).
But it's not their only priority when doing matchmaking they also have different criteria like making the waiting time shorts, puting players from same country together....
Basically it means the 50% value they want achieve is only theoretic due to these others criteria and due to the fact that their MMR models can't predict everything in a game. For example their system can't take account if a player is improving his MMR and will likely have a way higher MMR in the next 50 games.
So let's assume for every particular game the real match up odds is more likely to be often in the 30%-70% range than to be 50% acurately, meaning that if the same both team play 100 times one will win beetween 30%-70% of the times instead of this 50% ideal and theoretic value; which is actually quite a natural assumptions.
And now knowing than someone has plenty of data to analyse the two teams and can actually see if the match up is indeed exactly even or a bit one sided, predict 65% isn't that impressive at all; 85-90% would be impressive.
Quoted from the blog post:
"Current matchmaker criteria:
1. The teams are balanced. (Each team has a 50% chance to win.)"
Per your post - maybe they should change that line to 30%-70% chance to win?
Fearohnum
"d) There's plenty of things I plan to add besides DSR rating."
any eta <3?
@Lord russia
In the next few days - as DSR calculation will get stable :)
@ Fearohnum // MyDotaSkill.com
I don't think you understand basic statistics. "Each team has a 50% chance to win." It's a value they can only garantee on average, they can't garantee to put the odds for a particular game at 50%, but you try to predict wich team will win for a particular game.
Not only that but it's only one of their criteria.
Ok, they guarantee 50% chance to win on average.
I guarantee 65% chance to predict the winner on average.
Do you understand? Or it's still too hard?
Is there a "hidden" list to see more than top #20? I mean there is a "hidden" faq in /faq.php.
I played solo queue for my last 45 games, my rating suddenly dropped from 9xx to 4xx. I won some, I lost some, my performances were normal. I'd like to know finally where I am standing, but this rating won't tell me that. I sometimes checked people in my games and I had 1100-er guys and 200-er guys
I hope Valve doesn't just give us a rating but makes it visible, where you are actually standing. I don't even care a lot about other people, I want to know my rating and that mine is right. This thing tells me apparently not a lot.
Also that: http://mydotaskill.com/index.php?id=18035715 Owner is apparently god of Dota 2: 3594 points...
Maybe I need to wait until all scores are updated? oO Or it's just very bugged? oO
Your score is fresh and depicts the average opponent skill from past games.
Ok so it definitely doesn't work, I did not stack at all. Ok at least I know now.
Edit: Yes if I stacked I might be lower now I understood that. But I didn't do so or do you think I played every game in another stack? I even chose "Do not match with parties when alone". And you updated your comment, suddenly I did not stack in your data. Well that's true but how it comes you've seen it like this before. And your own score dropped in 10 minutes from 3500 to 650, wow.
Dsr is accurate if you are going to play last 20 games solo queue. When I played 20 games solo queue was 1550 Dsr,when I played last 10 with some friends I droped fast to 750 = ) .
@Captain
Well, I've checked and there's plenty of 48% winrate and 1.5 - 2 kda people in the matches. Also, the only team matchmaking rating I could find was 2500 ( which is quite low ).
But maybe you are an outlier and I'd have to investigate further.
@bogidoto
What he means to say is he's always 750, but he wants people to believe his rating is near mine.
I do not want to say I was at 900 or 400 or 1500 or 100, like I said I dont know where I am.
But I know: I did not stack for 45 games, I only played with equal or worse players before, I never played in team-matchmaking, I have no team (or do you mean when you stack with 1 or 2 it's team-matchmaking already?). So either matchmaking somehow throws me suddenly with much worse players in a game when I play at normal times in normal solo queue All-Pick or something else is wrong.
Lost more games than I won in the last week for sure, but my DSR went up 108 point from 771 to 879. I have been queuing up with better players but have also won some of those matches. i still play solo more than stacked however.
Emerald now. I thought that was top 3% or something... not sure. I remember it was platinum, emerald, pro, mythril
#hardwork #dedication
@Concede
You can check where you are here:
http://i.imgur.com/HPBoork.png
Mine just went form scrub to pro on my other acc
http://mydotaskill.com/index.php?id=http%3A%2F%2Fdotabuff.com%2Fplayers%2F134235086
thank god it got refreshed