EPTGeneral Discussion - Page 3277
| Forum Index > Dota 2 General |
IRC chatter should remain in IRC - http://webchat.quakenet.org/?channels=tl.dota2 Posts that relate to topics with their own thread, such as in News, Tournaments or Strategy should go in those threads. | ||
|
MrCon
France29748 Posts
| ||
|
RuskiPanda
United States2906 Posts
On July 09 2013 09:40 Zealos wrote: Yeah, eventually it just ignores my request :c It's trying to tell you to not fall into a dark lifestyle of being consumed by rares :o. | ||
|
Mithhaike
Singapore2759 Posts
On July 09 2013 11:10 MrCon wrote: We had several cases here of people smurfing and getting back in very high in less than 10 games. Obviously the people they got matched with will think that matchmaking is dumb and matched their 1000 wins account with a 8 wins account. When the 8 wins account is possibly better than them. From what i understand, the starting variation for new accounts are very large. If they do very well they can possibly end up quite high on the MM, however there's a set limit to it as I do not think it's possible for a under 10win account to end up on the first page. What happens is people's feeling gets hurt when they see they have a 10win account in the same game as their 1000games played account. There's no way I am on the same level as this newb they feel. It only mean 2 things. 1) They may have a thousand game, but if their still in Normal, it matters not thus they have a under 10win account guy with them. 2) They may be decent really, in very high or something. This may be a sign that the starting variance for new accounts are set too high, they should have to play for at least a set amount of games to be able to reach those levels. As for ranked/visible MMR, I am of the opinion that it should not be in DOTA2. Their only good for a general overview of people, but are extremely limiting in terms of scope & what that guy can bring to the table.For example you can definitely say a 1900rating guy is better than a 1500rating guy in overall terms of mechanics & skill, but exceptions can apply, for example that 1900 guy may have gotten so high only playing carry and cant support to save his soul, yet the 1500 guy is the best support in the world but his carry just can't win the game for him to raise his rating(known as ELO hell), resulting in that support guy having to change to play something he doesn't enjoy(carry) to be able to raise his rating. It also introduces too much undesirable elements to the games which can be abused(such as e-preen, forcing people to do things & griefing). God forbid if we end up like HoN where KDR is everything and your color determines which position you play. | ||
|
Rybka
United States836 Posts
Let's say they can even get within +/-10% in 50 games... once the matchmaking algorithm starts looking outside of a "perfect match," and in both directions, you could be looking at a really big range. | ||
|
canikizu
4860 Posts
On July 09 2013 11:10 MrCon wrote: We had several cases here of people smurfing and getting back in very high in less than 10 games. Obviously the people they got matched with will think that matchmaking is dumb and matched their 1000 wins account with a 8 wins account. When the 8 wins account is possibly better than them. It's all about ingame performance though. They got matched with a 8-win account, sure, but if those people perform excellently in game, I doubt no one will say any bad thing about that 8-win account. But in most case, the low-game-count- accounts usually don't perform up to the standard, that's why most people rage. I think a system shouldn't be designed to assume everybody is a smurf. The concept of smurfing, at core, still mean that you are trying to cheat the system. A system shouldn't assume everybody who has 8 win streak a smurf and accommodate by putting them through ranks unconditionally. It's gonna be hard to design a system to welcome and not overwhelm lucky newbies, while good enough to get smurfs up to their level conditionally. | ||
TheYango
United States47024 Posts
On July 09 2013 11:10 MrCon wrote: We had several cases here of people smurfing and getting back in very high in less than 10 games. Obviously the people they got matched with will think that matchmaking is dumb and matched their 1000 wins account with a 8 wins account. When the 8 wins account is possibly better than them. It's also better than the alternative. You're less likely to lose players by putting accounts with ridiculously high winrates in their first few games too high than you are by not pushing them up to higher MMRs fast enough. A 1000-win player is less likely to quit DotA over this than the newbies that the smurf plays with in his first few games. You wan't to quickly shunt the player out of the zone where he's queuing against new players if he's clearly a lot better than them, even if he has to lose a few games to come back down to where he's supposed to be. On July 09 2013 12:04 canikizu wrote: I think a system shouldn't be designed to assume everybody is a smurf. The concept of smurfing, at core, still mean that you are trying to cheat the system. A system shouldn't assume everybody who has 8 win streak a smurf and accommodate by putting them through ranks unconditionally. It's gonna be hard to design a system to welcome and not overwhelm lucky newbies, while good enough to get smurfs up to their level conditionally. The probability of a new player having an 8 game winstreak in their first 8 games is ridiculously low unless they stacked most/all of those games, and we've already established how the whole thing goes to shit when you start stacking with wide MMR disparities. A new player doesn't get matched 8 games at beginner skill level and then get instantly bumped up to Very High, that's not how this works. He gets matched in games of gradually increasing skill and starts dropping down when he starts losing some games. So if the new player has a 50% chance to win against newbies, by the 4th or 5th game, he's probably only got a 30-40% chance to win against players that are somewhat better than newbies if he's won 3-4 games straight. The chance of winning 3-4 more games against players better than you like this is really really low. You've got like a 1/5000 chance of a new player accidentally boosting himself into an MMR he has no way of handling. | ||
|
Caladbolg
2855 Posts
Of course this is anecdotal, but it's given me hope that if I start cranking out the games and get to my old skill level, the MMR will adjust and will keep giving me better and better opponents. The anomalies of MMR will principally be the result of stacking and intentional feeding/trolling, because smurfs will be easily shoved out of the lower tier games. | ||
|
canikizu
4860 Posts
On July 09 2013 12:22 TheYango wrote: The probability of a new player having an 8 game winstreak in their first 8 games is ridiculously low unless they stacked most/all of those games, and we've already established how the whole thing goes to shit when you start stacking with wide MMR disparities. A new player doesn't get matched 8 games at beginner skill level and then get instantly bumped up to Very High, that's not how this works. He gets matched in games of gradually increasing skill and starts dropping down when he starts losing some games. So if the new player has a 50% chance to win against newbies, by the 4th or 5th game, he's probably only got a 30-40% chance to win against players that are somewhat better than newbies if he's won 3-4 games straight. The chance of winning 3-4 more games against players better than you like this is really really low. You've got like a 1/5000 chance of a new player accidentally boosting himself into an MMR he has no way of handling. 8 win streak was clearly an example, and an extreme case, but the point still stands though. We've been seeing a lot of people who has 20, 30, 40 win get matched with people who have 300, 500, 1000 win, and the 20win guy ended up autoattacking creep, not buying any sustains, feeding in obvious situations,..v.v. We all know that if you give it time, that guy will end up at the right place, but the problem is the system jumps the gun too fast, and by jumping it too fast, it creates bad playing experience for the other four, or nine. Right now, I feel like the system is like this, for example: you start at 1, and the highest level is 100. Normal system: you jump level after a win, something like 1->5->10->15->20.....->100 or 1->10->20....->90->100 Current system: 1->50->75->88->94->97->98->99->100 Of course all of this is speculation. | ||
|
Blitzkrieg0
United States13132 Posts
On July 09 2013 13:14 canikizu wrote: 8 win streak was clearly an example, and an extreme case, but the point still stands though. We've been seeing a lot of people who has 20, 30, 40 win get matched with people who have 300, 500, 1000 win, and the 20win guy ended up autoattacking creep, not buying any sustains, feeding in obvious situations,..v.v. We all know that if you give it time, that guy will end up at the right place, but the problem is the system jumps the gun too fast, and by jumping it too fast, it creates bad playing experience for the other four, or nine. Right now, I feel like the system is like this, for example: you start at 1, and the highest level is 100. Normal system: you jump level after a win, something like 1->5->10->15->20.....->100 or 1->10->20....->90->100 Current system: 1->50->75->88->94->97->98->99->100 Of course all of this is speculation. Starting people at the very bottom of the barrel would alienate your playerbase since the average skill is going to be at 50. Starting everyone at 1 instead of 50 just means it would take longer for most people to reach their trueskill. Unless you have some method of determining the true skill of a player without them actually playing the game then somebody is always going to get shafted while you determine what it is. Outliers are not evidence that the system is broken and neither are players who have a small sample size of games to work with. MMR systems are designed to work over the long term. | ||
|
Esoterikk
Canada1256 Posts
| ||
|
anonymitylol
Canada4477 Posts
On July 09 2013 13:52 Esoterikk wrote: Anyone know why it default puts you in finding solo match until you cancel and requeue like 6 or 7 times| No clue, happening to me too | ||
|
r.Evo
Germany14081 Posts
On July 09 2013 13:14 canikizu wrote: 8 win streak was clearly an example, and an extreme case, but the point still stands though. We've been seeing a lot of people who has 20, 30, 40 win get matched with people who have 300, 500, 1000 win, and the 20win guy ended up autoattacking creep, not buying any sustains, feeding in obvious situations,..v.v. We all know that if you give it time, that guy will end up at the right place, but the problem is the system jumps the gun too fast, and by jumping it too fast, it creates bad playing experience for the other four, or nine. And how often do you check the amount of wins for all 9 people? Doesn't exactly sound hard to get confirmation bias into "seeing this alot" or more important "seeing this more often than the other extreme". | ||
|
Adrian_mx
Mexico1880 Posts
 bots keep going offline.. | ||
|
miyuki
Finland47 Posts
On July 09 2013 14:29 r.Evo wrote: And how often do you check the amount of wins for all 9 people? Doesn't exactly sound hard to get confirmation bias into "seeing this alot" or more important "seeing this more often than the other extreme". I think those low game count players are just stacked with a friend with loads of games played. Or you are solo queuing CM and the matchmaking doesn't match you with people with similiar MMR, simply because there is not enough players for that. | ||
|
BoesFX
1466 Posts
On July 09 2013 14:55 Adrian_mx wrote: well this sucks, dota2lounge isnt letting me bet on LGD.cn bots keep going offline.. Same situation with me. Is this a sign that I should not bet on LGD.cn? | ||
|
Itsmedudeman
United States19229 Posts
| ||
|
tapes
487 Posts
ELO hell doesn't exist. | ||
|
Kishin2
United States7534 Posts
| ||
|
icystorage
Jollibee19350 Posts
On July 09 2013 15:25 Kishin2 wrote: Is it me or does Dota 2 not actually have any great casters? Relative to sc2, at least. korean casters are the best even though you dont understand a single thing what they are saying | ||
|
rabidch
United States20289 Posts
| ||
| ||
