I really had no idea where I was at, I got into VH bracket a while ago (just a couple games before they stopped showing it), and almost immediately afterwards I went on a huge loss streak, and was in limbo since then. I was expecting something like 500 lol
Hmm let's put it this way. I've gotten better since DBR halted. DBR was obviously a more accurate measure than this system since it had far more information (a complete set...).
DBR: top 97-99% (I can't remember exactly)
This: bronze 80% (726)
hmm...
Creating a more accurate measure of a rating is not that hard... it's a partially observable problem (in terms of the players playing), and there is plenty of theory used to estimate partially observable states. It just takes someone with a Ph.D. to work on it rather than a dota player who jacks off all night and smokes weed.
@Vandal: you have to keep in mind that you didn't need to ask for a DBR. So a lot of bad people who didn't even check their dotabuffs got their DBR. On the other hand, you need to ask for your DSR. This means most people who got their score are good players simply because they care enough to ask it. So there are a lot of awesome players, rated over 1500, and few players under 900.
The percentile among the people who have a score. 99% of the players don't have one, and most of them are bad, while most players who asked for a score are at least decent.
ahahhaha my smurf acc in which i play when im banned and with completely fucking retards everywhere (frieds and shit too) has more rating than me
WTF IS THIS SHIT
look
Player number 126363354 DSR is:
1253
This player has a Platinum DSR score (top 10%).
compare it to my 774 u motherfucking pieceof shit!
FUCK YOURSELF AND YOUR FUCKING SYSTEM
@Sano,
I was also higher rating than most of my friends who are "decent". In this system, I am pretty much a bottom feeder.
And your point just makes it more transparent that the rating is completely flawed. The coder claims to account for the skill of your opponents, but how can he do that if he doesn't even assign a score to every player?
It's no coincidence that there is a large concentration of people posting about how their smurfs that are in a way lower MMR have a higher DSR than their main. It's because this website rewards stomps way more than anything else (and so it rewards stacking too more than anything else).
Yeah, which is pretty flawed. A couple of the best players I know have ~50% win rates, and another has a 53% win rate. The cause of their low WR is basically that they tend to play with IRL friends that aren't good.
that is odd. He might be taking the avg data from dotabuff "avg winrate of heroes" and compare it with your winrate so while his winrate might be high with wisp if you take all his 0% winrates in it brings him down. Do you know of any other players with high winrates and low dsr?
I just looked at players that had most of their games on just one hero. Pirate also has a 3 digit DSR even though he is a lot better than that. There sure is something weird going on in his algorithm.
Moving away from the abusers now.
http://dotabuff.com/players/86802544 - 76.27% (r1sk) 1449 - this is the one that confuses me. High win rate, good KDAs...captain of N9 and Chains Stack
my first thought was because there wouldn't be many high rated players in aus, maybe he plays lower rated players more often and hence has a relatively low rating himself, but balls and noob.you suggest otherwise
http://dotabuff.com/players/129606815 - 77.88% 1693
http://dotabuff.com/players/86809886 - 72% 1778
And yet, players like dendi/singsing, who you wouldn't expect to have a super high rating because of their playstyle in pubs, resulting in a relatively low win rate in comparison to most pros sit at around 1450 and 1533, both with ~55% win rate.
Its going to be heavily inaccurate for smurfs, but its worth noting that other rating systems work in the same way, although its not dealt with appropriately in this system. In HoN, your first 20 matches were much more sensitive to changes in rating, and you could jump from 1500 to 1700 in like 3 games or something (immortal). If you look at the player ladder, there are 3 players in the top 10 MMR with less than 300 wins each, gives you an idea that you don't need hundreds of wins to reach the top.
Onto another point. Remember back to DBR and remember the player spread? This is important when judging your rating. Not sure what the lowest possible was, lets just say 900. At about 1800, you were in the top 1 or 2%, and the highest players were like ~2600 or something. Understand that while you might only have 900 rating or something in DSR, that doesn't mean that you are that low at all, especially when you remember that unlike the DBR database which had every player, DSR is going to have much more skilled players on average, so being even at 600-800 isn't that bad either.
tl;dr
-inaccurate for smurfs, I'm guessing its unable to accurately evaluate the strength of allies/opponents and as a result, heavily inflates your rating
-your rating is better than you think, think back to DBR and also realise how much stronger the playerbase on DSR is. It may be that ~2000 (diamond) DBR is the same as ~1000 DSR or something.
-KDA/win rates are factors, most likely relative to the hero
-stacks are factored in
-quality of opponents are factored in, but not accurate due to missing players
Just one last point, realise that not all wins and losses matter. For those that played HoN, they would know that an individual match could be meaningless to your rating, while others could allow you to gain or lose heaps, ie. +5/-0.1 or +0.1/-5 or something. Due to the dud MM that we have, we can probably expect a lot of these, and even more so if you stack on a shitty server, where its more likely that you will have easier matches, while on SEA/US/EU, top stacks are able to play other top stacks quite often.
the worst part about this system is the person who created it misleads people to think skill level is calculated into the equation. There are alot of games where 40%-50% of the data is hidden and trying to make a analysis when almost half of the data is gone will almost always lead you to a false answer.
It doesn't feature in the skill of enemies or allies, so beginners can have the same skill rating as pro's. (since KDA and Windrate in very low games are the same as in very high games).
- Fixed a major bug causing many high-level players to receive low scores.
- Changed DSR calculation mode to "fresh". Now only your recent games count towards your ranking.
- Improved opponents skill approximation method. Increased resistance against smurfing and stacking.
- To get a higher score, avoid stacking with worse players for several games before checking.
- Changed mode of delivery from queue-based to instant ( until server can handle it).
- Rescaled the skill range to a more broad spectrum. New results are not comparable with old.
- If you have extremely low score and you feel you should do better, it's caused by your recent stacking with lower skilled players.
- DSR will now update daily.
Well, its recalculating, i already send some links in it, when its done im gona compare it =D
HEY DUDE Benao doto best doto I CHECKED YOUR PROFILE YOU HAVE NO HEROES WITH OVER 3 KDA EVEN IN MOST PLAYED.... WHY ARE U FUCKING COMPALINING YOU ARE TRASH? NOT EVEN ABOVE 3 KDA...
1118, how the hell does this even work.
FUCKING LAUGHABLE
Player number 78410961 DSR is:
774
This player has a Silver DSR score (top 60%).
I don't understand why people who stack are suprised that they have a high rating.
It really just seems to access your win %.
912, a lot higher than I thought!
I really had no idea where I was at, I got into VH bracket a while ago (just a couple games before they stopped showing it), and almost immediately afterwards I went on a huge loss streak, and was in limbo since then. I was expecting something like 500 lol
Hmm let's put it this way. I've gotten better since DBR halted. DBR was obviously a more accurate measure than this system since it had far more information (a complete set...).
DBR: top 97-99% (I can't remember exactly)
This: bronze 80% (726)
hmm...
Creating a more accurate measure of a rating is not that hard... it's a partially observable problem (in terms of the players playing), and there is plenty of theory used to estimate partially observable states. It just takes someone with a Ph.D. to work on it rather than a dota player who jacks off all night and smokes weed.
How is this suppposed to measure the skill of your opponents, when their profiles are hidden? MDS is flawed.
@Vandal: you have to keep in mind that you didn't need to ask for a DBR. So a lot of bad people who didn't even check their dotabuffs got their DBR. On the other hand, you need to ask for your DSR. This means most people who got their score are good players simply because they care enough to ask it. So there are a lot of awesome players, rated over 1500, and few players under 900.
They've been updating scores but mine still hasn't changed
That doesn't seem accurate at all Sano.
MDS tells you what percentile you are in. Im 1118 which is top 10% on this account.
The percentile among the people who have a score. 99% of the players don't have one, and most of them are bad, while most players who asked for a score are at least decent.
im at 1119 ahh yeee
ahahhaha my smurf acc in which i play when im banned and with completely fucking retards everywhere (frieds and shit too) has more rating than me
WTF IS THIS SHIT
look
Player number 126363354 DSR is:
1253
This player has a Platinum DSR score (top 10%).
compare it to my 774 u motherfucking pieceof shit!
FUCK YOURSELF AND YOUR FUCKING SYSTEM
@Sano,
I was also higher rating than most of my friends who are "decent". In this system, I am pretty much a bottom feeder.
And your point just makes it more transparent that the rating is completely flawed. The coder claims to account for the skill of your opponents, but how can he do that if he doesn't even assign a score to every player?
It's no coincidence that there is a large concentration of people posting about how their smurfs that are in a way lower MMR have a higher DSR than their main. It's because this website rewards stomps way more than anything else (and so it rewards stacking too more than anything else).
542 gg
relax Pringles dsr is scaling wr% + comparing % and games played with users on that homo site
Yeah, which is pretty flawed. A couple of the best players I know have ~50% win rates, and another has a 53% win rate. The cause of their low WR is basically that they tend to play with IRL friends that aren't good.
what is actually a good ranking because i have 1300 but my friends all have like 1500 :D
It's not working. My main has 640 ranking, my smurf has 1200. It's certainly bullshit rating.
So this just takes wr% into account? Useless.
benao , maybe it has something to do with your kda scores :S
Yup, they look at your wr, your KDA, your hero perfomance compared to other players and how good your opponents are.
It does not take into account who you fight I have done the testing on over 10 players(if it does it it VERY low representation)
http://dotabuff.com/players/107823071
fairly good kda but low w/l got him a 785 rating which is very low
winrate plays the biggest role in this guys algorithm the other sections are not even close
Could you share your testing?
Also, even people with insane WR and many games can have extremely low DSR, just look at this guy http://dotabuff.com/players/86736893
that is odd. He might be taking the avg data from dotabuff "avg winrate of heroes" and compare it with your winrate so while his winrate might be high with wisp if you take all his 0% winrates in it brings him down. Do you know of any other players with high winrates and low dsr?
For example if I take your data, you will score high because your hero pool is balanced out more
I just looked at players that had most of their games on just one hero. Pirate also has a 3 digit DSR even though he is a lot better than that. There sure is something weird going on in his algorithm.
It definitely looks at more than win rate..
http://dotabuff.com/players/145389281 - 90% win rate (abuser..) 1137
http://dotabuff.com/players/119829431 - 88% - 1419
Moving away from the abusers now.
http://dotabuff.com/players/86802544 - 76.27% (r1sk) 1449 - this is the one that confuses me. High win rate, good KDAs...captain of N9 and Chains Stack
my first thought was because there wouldn't be many high rated players in aus, maybe he plays lower rated players more often and hence has a relatively low rating himself, but balls and noob.you suggest otherwise
http://dotabuff.com/players/129606815 - 77.88% 1693
http://dotabuff.com/players/86809886 - 72% 1778
And yet, players like dendi/singsing, who you wouldn't expect to have a super high rating because of their playstyle in pubs, resulting in a relatively low win rate in comparison to most pros sit at around 1450 and 1533, both with ~55% win rate.
And looking at the two players with <50% win rate on this page, one a smurf (probably not), HOWDOYOUGETSLICED and Vawb, who both have pretty high KDAs
http://dotabuff.com/players/107823071 49% with ~1000 matches, 785
http://dotabuff.com/players/130682538 45% with ~160 matches, 912
Its going to be heavily inaccurate for smurfs, but its worth noting that other rating systems work in the same way, although its not dealt with appropriately in this system. In HoN, your first 20 matches were much more sensitive to changes in rating, and you could jump from 1500 to 1700 in like 3 games or something (immortal). If you look at the player ladder, there are 3 players in the top 10 MMR with less than 300 wins each, gives you an idea that you don't need hundreds of wins to reach the top.
Onto another point. Remember back to DBR and remember the player spread? This is important when judging your rating. Not sure what the lowest possible was, lets just say 900. At about 1800, you were in the top 1 or 2%, and the highest players were like ~2600 or something. Understand that while you might only have 900 rating or something in DSR, that doesn't mean that you are that low at all, especially when you remember that unlike the DBR database which had every player, DSR is going to have much more skilled players on average, so being even at 600-800 isn't that bad either.
tl;dr
-inaccurate for smurfs, I'm guessing its unable to accurately evaluate the strength of allies/opponents and as a result, heavily inflates your rating
-your rating is better than you think, think back to DBR and also realise how much stronger the playerbase on DSR is. It may be that ~2000 (diamond) DBR is the same as ~1000 DSR or something.
-KDA/win rates are factors, most likely relative to the hero
-stacks are factored in
-quality of opponents are factored in, but not accurate due to missing players
Just one last point, realise that not all wins and losses matter. For those that played HoN, they would know that an individual match could be meaningless to your rating, while others could allow you to gain or lose heaps, ie. +5/-0.1 or +0.1/-5 or something. Due to the dud MM that we have, we can probably expect a lot of these, and even more so if you stack on a shitty server, where its more likely that you will have easier matches, while on SEA/US/EU, top stacks are able to play other top stacks quite often.
obviously win rate isnt the only factor BUT it does play a huge factor into how they score in the this system
This system gives me a very low rating. The lowest I've seen anyone post. Therefore, I declare it to be super accurate.
WP sir, WP.
If you look at all of the high rated people in this system is that they have a good spread overall between all the heroes
Monkeh your posts make me laugh.
loled
"This player has a Beginner DSR score (top 99%) "
the worst part about this system is the person who created it misleads people to think skill level is calculated into the equation. There are alot of games where 40%-50% of the data is hidden and trying to make a analysis when almost half of the data is gone will almost always lead you to a false answer.
it is like trying to calculate 2 + X = Y and trying to figure out Y
I am 677 bronze tier scum. Time to go drown myself in alcohol and uninstall.
It doesn't feature in the skill of enemies or allies, so beginners can have the same skill rating as pro's. (since KDA and Windrate in very low games are the same as in very high games).
- Fixed a major bug causing many high-level players to receive low scores.
- Changed DSR calculation mode to "fresh". Now only your recent games count towards your ranking.
- Improved opponents skill approximation method. Increased resistance against smurfing and stacking.
- To get a higher score, avoid stacking with worse players for several games before checking.
- Changed mode of delivery from queue-based to instant ( until server can handle it).
- Rescaled the skill range to a more broad spectrum. New results are not comparable with old.
- If you have extremely low score and you feel you should do better, it's caused by your recent stacking with lower skilled players.
- DSR will now update daily.
Well, its recalculating, i already send some links in it, when its done im gona compare it =D
my # in que goes up over time lol
^
same here
If the programmer knew how to count, it would prolly be better
Man the queue is fucked :/
HEY DUDE Benao doto best doto I CHECKED YOUR PROFILE YOU HAVE NO HEROES WITH OVER 3 KDA EVEN IN MOST PLAYED.... WHY ARE U FUCKING COMPALINING YOU ARE TRASH? NOT EVEN ABOVE 3 KDA...
NOT EVEN ABOVE 3 KDA YOU ARE PURE GARBAGE.
ahahhaahhaha
ignorance at its best
guess its accurate now :$
>Benao doto best doto
1796
yup shut it down the best player is on the top now
This rating is so broken. I'm fucking terrible and I'm apparently 1641. What in the name of everloving what.
@CuntDota
I was about to say don't be so hard on yourself, and that I'm sure you are actually pretty good, but I just looked at your DB.
Yeah man, you are like my level kind of player, with fewer games. 1641? Wat?
teach me ur secrets
@Sampson
Go in to every game assuming the enemy sucks.
@sword
I get constantly put against tryhard 5man stacks all day every day every single game. Its fucking ridiculous.
@concede
This isn't true. I have a 49% winrate (I taught myself dota2 lost a ton early)
Now I solo queue against EG players / warden stacks etc... 1570ish mithril rating. So it definitely isn't based on winrate only.