exactly what i mean..
if you are a good player making smurf account and win 100 out of 150 matches
your rating will be 1400+
my point is this DSR thing can only calculate only if you play with and against people that has the same skill level as yours. but what if you don't ? then it will turn messy.
But take that good player and put him in his bracket and his DSR wont even reach 1400, and it takes more than 500 games to know where you supposed to be.
@murs, he keeps improving the system. If you look at the top of the list, which was mostly 1900+, now its only people who are 1700+. Your rating relative to others is the same.
@sano - your friend stacks heavily - possibly with much worse players and it drags his score down. If you stack with anyone, your score will converge towards his/her score and his/her score will converge towards yours.
^Wow, stacking results in winning? Winning results in a higher rating? you don't say... And not to mention you have won 19 of your last 20 games by stacking.
I think its funny that being in the top 5% is ridiculously horrible.
most interesting part of this is everything you have said and presented implies that once enough games have been played, a player's skill bracket will probably never change. It implies that its better to start a new account than try to improve an existing account.
this system (admittedly) rewards you with a higher rating by playing with your higher rated friends and punishes you by playing with your lower rated friends. only if you always solo queue does your rating really represent your skill level. because the system basically just measures the average skill level of each game you play in and applies that to you, instead of measuring your actual skill level, which does not go down just because you are slamming an easy pub game with your crap friends.
the elo system that dotabuff had did a much better job of classifying players in this way. i play with one friend who's not that great and in dotabuffs system I was top 1% and he was bottom 25%. now, since we basically queue up together every game, our DSR's are identical and it thinks we are both top 10% players, even tho there is a massive skill disparity between us.
now i could make my DSR go back up to top 1% just by playing 20 games in solo queue. hope this example is clear enough that the system doesn't work so well in its current iteration.
the best solution would be to make a pure win/loss elo system like dotabuff had, and just approximate the rating of "Anonymous" players based on the average rating of the rest of the team. there is some error there but in the end, and especially in the high level games where majority of players have privacy option disabled, you get pretty good data and can rank players much better than some other secret formula that makes obvious mistakes.
Yesterday i was able to see my DSR and now for some reason I can't. I type my ID and it answers "Player number 101835435 skill rating is still being calculated. Currently, your position in the queue is ."(no position number). How to fix this issue? Or just tell me my current DSR for now. Thanks.
Hello MyDotaSkill.com #650 DSR, having high DSR >1700 doesn't place you into very high makemaking bracket? or even in top live pages 1-5 ? ; or it depends on how many games you've played?
its silly that by losing games in a higher bracket you can make your DSR increase and that by winning in a lower bracket you can make your DSR decrease, just by choosing who you queue with, right? shouldn't a win always be to your benefit and a loss to your detriment?
@mydotaskill: You can have 58% winrate and play against bad players..LOL ~2500 games and always in very high bracket http://pubstats.me/116919366/ how the fuck i can mmr with bad player. My account ussualy on top live page 1-4
I will be honest and just say im lazy and dont feel like reading through the pages. Is it better to solo queue than play with 1 more person or more? To improve rating.
MyDotaSkill #650 -> queuing for mm alone PUNISHES me; check my recent matches. I am queued with people with lower skill than I have because they are in a stack and I am alone so their MMRs combine to make a higher MMR that matches mine. So it seems like your system will look at these low skill players and say look at LitBull he is queuing with lower skill players...
Unless I'm understanding it wrong; however you can't know if people are grouped together or not can you?
^
if you are queuing alone, why aren't you selecting solo queue only?
Edit: @head&shoulders The skill rating being changed by queuing up with friends of a lower rating will have NO effect on your actual dota rating. Winning is better than losing, even if you gain nothing by winning its still better than a loss.
If you want to inflate your skill ranking, convince better players to queue with you and win games.
regardless, this skill rating has absolutely no effect on your in game rating
I've been playing captain's draft mostly because AP is a shitfest and it already takes ~4-5 minutes to grab a game in CD. Solo queue greatly increases wait times. I use this account to queue alone for MM and a smurf to play with lower / equal skilled friends.
Something just occurred to me. 90% of the players in my game a marked anonymous by dotabuff. That implies they are not sharing their information with 3rd party sites like dotabuff.
If that's the case, and the DSR from mydotastats is based on the ratings of the players I play with, how is it accurate? In high level matches where the top 1% play, most players share their information, but for the other 98% of players, how is this accurate is the system can't determine the skill level of the the vast majority of players I play against.
Does mydotastats have access to the API information of players who don't share their information? Will the site only work properly for players who are in games where the majority of players share their information?
The latest update looks very good from what I have seen thus far, still disappointing to see you cop so much shit from it from butthurt kids on here and on reddit.
@skynet and yet i'm supposedly top 5%. Assuming skill level based on players not sharing their API is pointless. Taking a look at your dotabuff and DSR, do you think its fair to refer to you, and everyone in your games as noobs, most of the people in your recent games don't have dotabuffs, and you are gold ranked on the DSR. Im not going to assume you are fresh/bad because of it.
IT also doesn't answer the question I asked. Does this website give an accurate rating if the vast majority of the players in matches aren't sharing their information?
I still want to know how the ratings are calculated when 7-9 our of 10 players in matches aren't sharing their information with 3rd party sites.
You can go ahead and tell us why it appears to give an accurate (roughly what you would expect) in most cases then? The only strange ones that I have seen are where its a player who plays only in stacks, and hence may get an inflated/deflated rating OR people with extremely old accounts like this one which is pre-6.79, and more importantly, before they implemented stronger anti-smurf measures, which is why more recent smurfs have significantly higher ratings, not because they have good stats, but because they have actually played in much higher rating games than this account.
edit: and you heavily overestimate the average player if you think top 5% is high..
First off, "appears accurate" has been based on players that people know, aka pro players or highly skilled members of the community.
Second off, pretty much "roughly what you would expect" has been based of the top 5%, mostly the top 1% and the bottom 80% is almost completely ignored. I have not seen much if any gageable differentiation between players in the 70 vs the 30th percentile for example.
Thirdly, I asked a question if the system can work for players below the 85th percentile.
Lastly, no, i'm not going to "go ahead" and tell you anything. I asked a question, and I have no information shared with me regarding how mydotaskill is using the API. Why would i "go ahead and tell you" anything when those are the circumstances.
Edit: and no, i didn't estimate anything. I just don't make assumptions.
As Valve re-calibrated the skill brackets in the "First Blood" patch, Normal Bracket is now the bottom 50% of players, Very High top 25% and High Bracket in between.
I am currently at 1468 DSR... I guess that is just short of "mirthril", somewhere between top 0.1% and top 0.05%. I see many players in my games are similarly rated "jadeite" or "mithiril".
Looking over the last 20 games, 151 of 200 players had stats on... about 75% on. Probably that is not true for the typical dota player.
Will you share your strategy for accounting for the missing information?
@relentless, I can't be sure, but I think your numbers are wrong. Its really difficult to say because there are so many bad players out there, and that should be the end of it, but is it possible that VH players play more on average than say, high or normal bracket players? I don't mean in total, I mean each individual player in that bracket, on average. So while there might be 25% of the matches labelled as VH, there might be less than 25% of the player pool that play in the VH bracket.
And yeah, how he is getting data for missing players is really weird.
Also Relentless your score is very high cos you stacked with better players than yourself last 20 games. Chris and more ppl.Try solo queue 20 and than that`s real score. You ll be like 1100 max I think.
Good for you Bogi. Incidentally, I happen to have a far higher score than Chris or anyone in his stack. I don't claim to be better than they are... but I do I have much higher DSR. About half of my last 20 games are solo que.
@ Terrible, probably higher MMR players play more games but it is the only way to estimate the range of the skill brackets since Valve will not tell us. There was a dramatic change in the % of games in each bracket at the First Blood patch. Previously about 85% of games were in Normal bracket and less than 3% were in Very High Bracket.
Generally if players have stats off they are nub/lower skill. But its not always true. Some strategy much be employed to account for the missing information. If he is willing to share it we can discuss the merits. Much less data is missing for top rated players, but a lot is missing for lower ranked players. And of course for someone like Terrible who has had stats hidden for half a year... this can also distort a rating.
@way2high
exactly what i mean..
if you are a good player making smurf account and win 100 out of 150 matches
your rating will be 1400+
my point is this DSR thing can only calculate only if you play with and against people that has the same skill level as yours. but what if you don't ? then it will turn messy.
But take that good player and put him in his bracket and his DSR wont even reach 1400, and it takes more than 500 games to know where you supposed to be.
What's the "new model rollout" about?
Newer scores are lower on average by 200 than the scores from last algorithm. You are still near the top.
New model ratings aren't comparable to old model ratings. Don't throw such a hissy fit it makes you look pathetic. Good job you're good at the game.
what goes into these scores anyways
@murs, he keeps improving the system. If you look at the top of the list, which was mostly 1900+, now its only people who are 1700+. Your rating relative to others is the same.
also thank u for responding to me twice
@MyDotaSkill.com #650 DSR
Can you please tell me why my friend http://dotabuff.com/players/86901513 has silver DSR and I have pro? I think we're pretty much on the same level.
went from 262 to 534 in 2 day or 9 games. Not bad?
skynet during those 2 days the algorithm changed like twice.... don't compare new dsr to old
@sano - your friend stacks heavily - possibly with much worse players and it drags his score down. If you stack with anyone, your score will converge towards his/her score and his/her score will converge towards yours.
I knew my soloqueue endeavours would pay off. Every stacker I see has like 800-1200 rating.
^Wow, stacking results in winning? Winning results in a higher rating? you don't say... And not to mention you have won 19 of your last 20 games by stacking.
I think its funny that being in the top 5% is ridiculously horrible.
the other 95% must be a shitpool of despair.
I'm pretty sure I've played a total of 5 games out of my last 200 in a group with more than 2 players in.
Can you please tell me why my friend http://dotabuff.com/players/103340954 has a Mithril DSR score (1728) and me http://mydotaskill.com/index.php?id=http%3A%2F%2Fdotabuff.com%2Fplayers%2F116919366 but his winrate sucks
my winrate is 57.62% ( DSR score 1581) , him 52.08% ( DSR score 1728)
Winrate is a noisy metric. You can have 58% winrate and play against bad players.
In your case, your score may be lowered by stacking with worse players. Your score will converge towards the scores of people you stack with.
@klaze Apparently you think solo queuing and queuing with another person are the same thing.
@mydotaskill
most interesting part of this is everything you have said and presented implies that once enough games have been played, a player's skill bracket will probably never change. It implies that its better to start a new account than try to improve an existing account.
this system (admittedly) rewards you with a higher rating by playing with your higher rated friends and punishes you by playing with your lower rated friends. only if you always solo queue does your rating really represent your skill level. because the system basically just measures the average skill level of each game you play in and applies that to you, instead of measuring your actual skill level, which does not go down just because you are slamming an easy pub game with your crap friends.
the elo system that dotabuff had did a much better job of classifying players in this way. i play with one friend who's not that great and in dotabuffs system I was top 1% and he was bottom 25%. now, since we basically queue up together every game, our DSR's are identical and it thinks we are both top 10% players, even tho there is a massive skill disparity between us.
now i could make my DSR go back up to top 1% just by playing 20 games in solo queue. hope this example is clear enough that the system doesn't work so well in its current iteration.
the best solution would be to make a pure win/loss elo system like dotabuff had, and just approximate the rating of "Anonymous" players based on the average rating of the rest of the team. there is some error there but in the end, and especially in the high level games where majority of players have privacy option disabled, you get pretty good data and can rank players much better than some other secret formula that makes obvious mistakes.
Yesterday i was able to see my DSR and now for some reason I can't. I type my ID and it answers "Player number 101835435 skill rating is still being calculated. Currently, your position in the queue is ."(no position number). How to fix this issue? Or just tell me my current DSR for now. Thanks.
@SmallStepsCorrupt
There's a bug with some ID's - I will resolve it soon.
Your current DSR is: 1536
@MyDotaSkill.com #650 DSR
I couldn't add you back as friend, could you add me? :(
Hello MyDotaSkill.com #650 DSR, having high DSR >1700 doesn't place you into very high makemaking bracket? or even in top live pages 1-5 ? ; or it depends on how many games you've played?
its silly that by losing games in a higher bracket you can make your DSR increase and that by winning in a lower bracket you can make your DSR decrease, just by choosing who you queue with, right? shouldn't a win always be to your benefit and a loss to your detriment?
@mydotaskill: You can have 58% winrate and play against bad players..LOL ~2500 games and always in very high bracket http://pubstats.me/116919366/ how the fuck i can mmr with bad player. My account ussualy on top live page 1-4
http://dotabuff.com/players/120618402
Would you call this player to have the same skill as you?
@MyDotaSkill.com #650 DSR, Can you do mine plx. ID: Head & Shoulders
@Head and shoulders
Your DSR is: 1419
Just like to say w/e ur doing it looks better.
I will be honest and just say im lazy and dont feel like reading through the pages. Is it better to solo queue than play with 1 more person or more? To improve rating.
MyDotaSkill #650 -> queuing for mm alone PUNISHES me; check my recent matches. I am queued with people with lower skill than I have because they are in a stack and I am alone so their MMRs combine to make a higher MMR that matches mine. So it seems like your system will look at these low skill players and say look at LitBull he is queuing with lower skill players...
Unless I'm understanding it wrong; however you can't know if people are grouped together or not can you?
^
if you are queuing alone, why aren't you selecting solo queue only?
Edit: @head&shoulders The skill rating being changed by queuing up with friends of a lower rating will have NO effect on your actual dota rating. Winning is better than losing, even if you gain nothing by winning its still better than a loss.
If you want to inflate your skill ranking, convince better players to queue with you and win games.
regardless, this skill rating has absolutely no effect on your in game rating
@concede thx... also, I didnt even know u could find a match in solo queue :3. Last time i tried it took like 10+ minutes and it got declined
I've been playing captain's draft mostly because AP is a shitfest and it already takes ~4-5 minutes to grab a game in CD. Solo queue greatly increases wait times. I use this account to queue alone for MM and a smurf to play with lower / equal skilled friends.
Can anyone check my DSR please? i allowed dota 2 to share match history and searched my name for DSR but it gave me this>
"Error: Player number 0141613697 game data couldn't be fetched. Please try again soon."
http://dotabuff.com/players/83227395
Something just occurred to me. 90% of the players in my game a marked anonymous by dotabuff. That implies they are not sharing their information with 3rd party sites like dotabuff.
If that's the case, and the DSR from mydotastats is based on the ratings of the players I play with, how is it accurate? In high level matches where the top 1% play, most players share their information, but for the other 98% of players, how is this accurate is the system can't determine the skill level of the the vast majority of players I play against.
Does mydotastats have access to the API information of players who don't share their information? Will the site only work properly for players who are in games where the majority of players share their information?
Your score is: 381
I love this new rating system. seems to fill a vacuum where something should have been~
The latest update looks very good from what I have seen thus far, still disappointing to see you cop so much shit from it from butthurt kids on here and on reddit.
90% of them private = assume fresh noobs and/or bads.
Why were the elo values changed on some brackets?
@skynet and yet i'm supposedly top 5%. Assuming skill level based on players not sharing their API is pointless. Taking a look at your dotabuff and DSR, do you think its fair to refer to you, and everyone in your games as noobs, most of the people in your recent games don't have dotabuffs, and you are gold ranked on the DSR. Im not going to assume you are fresh/bad because of it.
IT also doesn't answer the question I asked. Does this website give an accurate rating if the vast majority of the players in matches aren't sharing their information?
I still want to know how the ratings are calculated when 7-9 our of 10 players in matches aren't sharing their information with 3rd party sites.
You can go ahead and tell us why it appears to give an accurate (roughly what you would expect) in most cases then? The only strange ones that I have seen are where its a player who plays only in stacks, and hence may get an inflated/deflated rating OR people with extremely old accounts like this one which is pre-6.79, and more importantly, before they implemented stronger anti-smurf measures, which is why more recent smurfs have significantly higher ratings, not because they have good stats, but because they have actually played in much higher rating games than this account.
edit: and you heavily overestimate the average player if you think top 5% is high..
First off, "appears accurate" has been based on players that people know, aka pro players or highly skilled members of the community.
Second off, pretty much "roughly what you would expect" has been based of the top 5%, mostly the top 1% and the bottom 80% is almost completely ignored. I have not seen much if any gageable differentiation between players in the 70 vs the 30th percentile for example.
Thirdly, I asked a question if the system can work for players below the 85th percentile.
Lastly, no, i'm not going to "go ahead" and tell you anything. I asked a question, and I have no information shared with me regarding how mydotaskill is using the API. Why would i "go ahead and tell you" anything when those are the circumstances.
Edit: and no, i didn't estimate anything. I just don't make assumptions.
As Valve re-calibrated the skill brackets in the "First Blood" patch, Normal Bracket is now the bottom 50% of players, Very High top 25% and High Bracket in between.
I am currently at 1468 DSR... I guess that is just short of "mirthril", somewhere between top 0.1% and top 0.05%. I see many players in my games are similarly rated "jadeite" or "mithiril".
Looking over the last 20 games, 151 of 200 players had stats on... about 75% on. Probably that is not true for the typical dota player.
Will you share your strategy for accounting for the missing information?
BogiDotO
Player number 134203835 DSR is:
1541
This player has a Mithril DSR score (top 0.05%).
Better than Relentless , now I can go sleep.
@relentless, I can't be sure, but I think your numbers are wrong. Its really difficult to say because there are so many bad players out there, and that should be the end of it, but is it possible that VH players play more on average than say, high or normal bracket players? I don't mean in total, I mean each individual player in that bracket, on average. So while there might be 25% of the matches labelled as VH, there might be less than 25% of the player pool that play in the VH bracket.
And yeah, how he is getting data for missing players is really weird.
Also Relentless your score is very high cos you stacked with better players than yourself last 20 games. Chris and more ppl.Try solo queue 20 and than that`s real score. You ll be like 1100 max I think.
Good for you Bogi. Incidentally, I happen to have a far higher score than Chris or anyone in his stack. I don't claim to be better than they are... but I do I have much higher DSR. About half of my last 20 games are solo que.
@ Terrible, probably higher MMR players play more games but it is the only way to estimate the range of the skill brackets since Valve will not tell us. There was a dramatic change in the % of games in each bracket at the First Blood patch. Previously about 85% of games were in Normal bracket and less than 3% were in Very High Bracket.
Generally if players have stats off they are nub/lower skill. But its not always true. Some strategy much be employed to account for the missing information. If he is willing to share it we can discuss the merits. Much less data is missing for top rated players, but a lot is missing for lower ranked players. And of course for someone like Terrible who has had stats hidden for half a year... this can also distort a rating.