what's with the constant need for visible ratings, you lika to sucka hard on ya' own e-peen dawg?
Back in dota 1,I would view professional players as Gods of the game. Having no rankings at all, I would be playing against normal pubs and thrashing them yet not sure where I stand in the general populace. However with Dota 2, I would get to play against professionals which gave me a gauge of where I am in the general populace.
But how do you know where you stand? Statistics don't really mean much as MM is calculated by valve's hidden rating.
The bracket standard deviation is huge.
I don't get why Valve isn't doing a standard ELO system like league. An ELO system does not work well with a team game like Dota, but at least it will have visible numbers and you can compare everyone. I mean, it wouldn't be that hard to implement an ELO rating right now. Either they are either developing a new, (more accurate) ELO-like rating system and a ladder and they aren't done OR they are actually retarded and nothing like that is in the plan for a long time.
But rest assured, I am sure there will be a day where we get ratings behind our names and a ranking ladder system. I mean, they have to......right? (I dislike League, but look at them, they are using ELO and they are doing fine)
@llVers
Valve does use ELO-type system.
Quote from Dota2 dev team:
Yes, we use a Elo-type system. We have data that shows that it correlates well with a number of quality metrics.
Showing players their Elo has proven to cause a lot of unwanted behaviour. (The same problems exist in Chess.)
We know that there really is value in knowing how the match was formed, and even in knowing which players were the best, and also who was partied up with whom. It helps to analyze your own skill and try to understand how to get better. We are actively working to try to figure out what data we CAN provide to players (perhaps after the match) to help them do this analysis, without causing too many of the undesirable effects I mentioned. Right now, if there's a steamroll, people look at the only data available to them, which is totally understandable. Also, I believe that there have been problems in the past where it was not doing a good enough job segregating truly inexperienced players from experienced players. We've improved on that. However, our definition of "new player" probably doesn't match exactly with many players. (Their definition is often: "the other guy has fewer wins that I do.") And we have to deal with smurfs and players coming in from other games in the genre, where win count is definitely totally unrelated to the skill level of the player.
"Elo is a TERRIBLE way to give players a sense of "progress." Many (most?) people reach a plateau, and their Elo stabilizes. It is simply not mathematically possible for Elo to keep increasing in general for players indefinitely as more and more games are played.
Given this reality, if players used Elo to measure "progress", we would constantly be reminding them that they are NOT making any. That would be really bad."
@adivity
-Yes, we use a Elo-type system. We have data that shows that it correlates well with a number of quality metrics.
What metrics? How strong is the correlation? What are the alternatives to ELO? How did they do on those metrics? BASICALLY, valve said nothing here in terms of actual information.
-Showing players their Elo has proven to cause a lot of unwanted behaviour. (The same problems exist in Chess.)
What was the unwanted behavior? If it's so bad, why can't you just tell me what it was rather than define it as bad for me? Surely, I'd conclude it was bad too, and your point would be stronger. Where is the evidence of causality? Where is the evidence this happens in chess? Did it happen in online chess or offline chess? If only the latter, how can we be sure it would happen in online DotA? How can we be sure it'd happen in this team game since chess is 1 v 1? BASICALLY, valve said nothing here in terms of actual information.
- big quote
He says there is value in knowing ratings and so forth. I will let it slide that he doesn't actually define those values since he was most likely quoting a person who stated the value. Hence, his statement isn't omitting information if we saw the complete conversation. However, he continues to say they work to show information "without causing too many of the undesirable effects I mentioned." Where did you mention them? What were they? So far, all I can see is the "undesirable effects" are those that were "unwanted behavior ... [that] exist in chess." Great, case closed. BASICALLY, valve said nothing here in terms of actual information.
-progress quote
This statement presupposes the only reason you'd be transparent with ratings is to convey a sense of progress and that people play games solely to make progress. And that's the reason it's a load of bullshit - both presuppositions are quite false in general.
1.) The value is CREATING progress, not SEEING progress. Transparent ratings would allow lesser players to gain a sense of humility and finally listen to someone who is statistically 10x the player he is. He'd then learn. If his ELO stays constant (since it is a "mathematical impossibility" for all ratings to increase indefinitely) so be it. We would still have more competitive, competent, and fun games that are more removed from high variance, tosses of the coin in terms of which team wins.
2.) People don't play games to see progress. They play games to have those moments where they do really, really well, and they play to enjoy a fair, competitive game where they duke it out. It's simply fun. As long as they are matched against fairly skilled opponents, you don't need to worry if a sense of progress is quashed through statistical data. Fairly matched opponents allow for domination here and there. It also allows for them to have fun and appreciate seeing that their matches were fair numerically.
3.) On the other hand, there is no way for someone to teach another player without insulting him - he will automatically assume he is better than you. There is no way for people to know if their games were fair - they will assume either that "MM sucks" or that "MM had to make an unfair match up to lower waiting times."
I'd like to conclude that valve's intellectual skill has been QUITE disappointing from the tidbits of algorithms I have seen them use. For example: Using a moving time window in which to count the number of reports, punishing if that number surpasses a constant threshold (3, for example, over a 1 week window) is the most immature, lazy, and uninspired algorithm I have ever seen. I just can't think of which college graduate of the highly mathematical CS degree didn't realize how that places a tremendous skew on bans, making bans highly favor people who play more than those who play less. I lost all faith in their ability to think critically/rationally/algorithmically when I read that was one of their algorithms. They have since changed the algorithm to a much better variety, counting the # of reports per game (to normalize for how much you play). Ok, great. They're not thinking like juniors in college. Let's see what else they can do. And of course, I do not trust them to tackle the highly nontrivial, extremely mathematical topic of ranking players. Why bring all of this up? Maybe someone will point out the obvious flaws in their shitty algorithm if they publicize it.
@Vandal
If you have a solution then head over to dev.dota2.com and enlighten Dota2 developers.
Valve does not use an ELO type system. They use a TSR type system. Really the fact that in this quote and others from the dev forums reveals is...clearly the people running it don't even know the proper terminology for the mathematical system they are using.
I have found every statement from Valve employees about their matchmaking system disillusioning because they reveal that they don't understand what they are talking about.
If only pesky devs understood how their creation works.
http://dev.dota2.com/showthread.php?t=98311&page=3&p=664107#post664107
Good thing you're ahead of them Relentless.
Also, I like how you provide enough sauce to support your claim.
the only sauce mr relentless has is the sauce dripping outta his butthole after gettin rammed and rekt
![[Lk].Zano](https://steamcdn-a.akamaihd.net/steamcommunity/public/images/avatars/b9/b98e01ce6ef4a5e926a44cdd1564d0e8e5fbfa36_full.jpg "[Lk].Zano")
For better or worse, Relentless is partially correct. If you understood what the TrueSkill system is, you could immediately see how it applies, and also be able to match up the programmatic inputs to the system. From what I have seen (which is only slightly more than what most people know), it is a heavily modified TrueSkill variant (makes sense since M$ did not release the complete spec, and require licensing agreements before they share it).
It is difficult to see now...since they have hidden more and more aspects of MMR that could earlier be used to show how it works.
One of the foundational assumptions of the ELO system is that the skill of a player can only change slowly. Because of this the equations that govern ELO systems set the change in rating from a given match as simply proportional to difference between the expected and actual performance.
TSR systems however are intended to be more flexible (like DBR which was also a TSR system). A TSR system is designed to converge quickly and be able to rapidly adjust ratings to changes in performance. To accomplish this it employees a second measurement that it intended to describe the uncertainty in each players score and allow more rapid changes when uncertainty is high and slow changes when it is low.
Before everything was hidden, even before the anti-smurf system was added...it was clear that MMR changed very quickly with win and loss streaks. This is a feature of TSR type systems and not a feature of ELO type systems. Consequently I know that MMR is TSR based not ELO based....regardless of what incoherent comments are made by the developers.
Signature Man (above poster) is correct. This is a heavily modified TSR type system. They put in some...I would say clumsy, additional algorithms for the anti-smurf and other extras.
Also while the patented super special extras are not free...you don't need them.
http://research.microsoft.com/en-us/projects/trueskill/details.aspx <----all you need to make one is here.
They probably started out with an Elo-type system and then tweaked it over time to take care of smurfing. In that way his statement makes sense even though it is incorrect.
Simple sollution for your problem GET A CLAN! And yes you perhaps will get one if you act nice and dont say "i have 1700 mmr in hon i be carry you sappart mai wayne!"
dota 1 is different from dota 2 since one, the overall skill lvl has gone up since more ppl understand mobas a lot more. and 2, you get placed with ppl the mms thinks you should be playing. as much as ppl don't wanna believe it isn't just your kda that gets you into other brackets, pretty much look at the record page and some of the stats shown there are also an indicator on what you need. if you are averaging around 100cs/20mins that's alright it for sure beats the overall dota 2 average around 80ish i believe. skill is based off an average in this game and how far you stray from it will either bump you up or down. it also depends on how you play your pool of heroes because that is also a factor. the reason why smurfs have an easier chance to bounce into the higher brackets is because they have no skill range yet. for instance a player with 500 wins plays and plays exceedingly well (compared to most of their games) they might walk away with 20mmr points at the end of the game but a smurf who does exceedingly well in their game might walk away with 1500-2000 mmr points because the system doesn't have a fixed bracket for it yet and isn't sure where the player should be placed
I'm sorry that annoys you. Yes, Elo is a man's name. I did not mean to give the impression it was an acronym. I just wanted to emphasize it so it stood out when reading.
Будь ласка, увійдіть до системи для написання коментарів.
Back in Dota 1 days, you could more or less tell how good you were against the general populace since you just win waaaay more.
But now, because of matchmaking which usually ends up in 50% wr, most players can't really gauge where they stand. Yes, there were visible brackets before and yes, there are stats available... but so what?
Brackets meant little because you had no idea where you stood in that bracket. In other words, its a really rough estimate which we don't even know how its calculated. You have good stats, so what? You could have 53% wr, decent KDA and be at the bottom 30% of the entire player base. You could have a negative wr, lacklustre KDA but play in high level games in the top 10%.
In other words, players often decide their own skill level based on their own opinion. You decide it through comparing yourself with the players you matchmake with, you decide it by watching competitive videos and asking if you could do that, etc etc. This often leads to bias and the dunning-kruger effect.
With no visual rating like a ELO or MMR, players are left to their own imaginations on where they stand. Some people feel they're much better than they really are and some feel they're bad when they're really capable.