@Hobbes@BestProfileName You misunderstand. I was not offering any kind of conclusion. I made a simple statement that I thought might add to the conversation. Did you know that fades have claws for arms?
Sure, comp players don't always try as hard as they could. I also see players with extremely high mechanical skill with a lower hive skill than I would expect. These same players make very poor decisions. Wob has the highest hive skill by far, but I doubt he is the most gifted player to ever grace NS2. Wob is very skilled, but he just has the right combination of skills that allow him to win often in pubs.
I don't see how my simple statement is in any way a counter to that. I guess it was taken the wrong way.
@Hobbes@BestProfileName You misunderstand. I was not offering any kind of conclusion. I made a simple statement that I thought might add to the conversation. Did you know that fades have claws for arms?
Sure, comp players don't always try as hard as they could. I also see players with extremely high mechanical skill with a lower hive skill than I would expect. These same players make very poor decisions. Wob has the highest hive skill by far, but I doubt he is the most gifted player to ever grace NS2. Wob is very skilled, but he just has the right combination of skills that allow him to win often in pubs.
I don't see how my simple statement is in any way a counter to that. I guess it was taken the wrong way.
I thought the people in the conversation were already likely aware of it, so took it as read. Fair enough if they were not.
Wob doesn't play many gathers and has barely played any competitive play for quite some time. He also tries really hard (nothing wrong with that). I've watched some other comp people (ADHD/Schu/asdf) for example. They gather and play a lot of comp, and generally don't try hard in pubs.
i play better when smurfing, averaging 10:1 ratio for kills and 40% accuracy, i actually only smurf while friends are around, good times.
is this because you're on rookie servers or is your smurf account seasoned?
btw, kind of unwise to go around with your name as 'ns2cheats.ru' or w/e your name was last i saw you in game, dont you think?
i only smurf anymore when I'm tryin to log into main steam accnt and it asks for steamguard mobile authentication code and I'm not near my phone. i havent set up steamguard on my alt accnt so i dont need my phone glued to me to get into this one.
i play better when smurfing, averaging 10:1 ratio for kills and 40% accuracy, i actually only smurf while friends are around, good times.
is this because you're on rookie servers or is your smurf account seasoned?
btw, kind of unwise to go around with your name as 'ns2cheats.ru' or w/e your name was last i saw you in game, dont you think?
I played fair on the rookie server, one-two rounds then i moved on and left and then went to [TA] i only play, some good players once in a while on [TA] however most people are greens who have evolved, and what does "seasoned smurf account" even mean?
I think what i've realised is, if you just don't interact with people and enjoy the game it's far easier, especially when you have people targeting you even when you are smurfing, peoples names come to mind it's like they want to bitch about smurfs because they aren't even good at the game, even when you aren't stacking they still have a little cry...
Regardless ratio is probably closer to 9:1 though varies.
In regards to unwise, why does my name have any affect on you or anybody else? anyone that is going to stick around already knows whom eachother is anyway so it's not likely anyone else other then a few new people are going to have a moan and that is not the case, besides the server owners know who we are anyway so there's that. Hell i have a database of smurf accounts from the Australian community from when i was maintaining our servers, you'd be surprised at the amount of people who pretend like they've quit the game who are actually still around.
@Archie you should "play" better with a smurf account if the hive score is lower. Hive is trying to balance and thinks your not as good, so it gives you easier opponents.... these easier opponents run in strainer lines and miss more bites, so you accuracy and life expectancy increase.
@Hobbeson as @Nordic had previously mentioned there is "mixing" what I was commenting on is that the "mixing" effect fails when people don't put for the the same effort on all servers. Specifically comp players who might not care as much about winning on Pub. Due to Hive ELO only looking at W/L the case competitive player base, hive score is lowered.
Granted that as long as the comp player puts forth a lower level of effort then hive score will still balance w/l. However, if a comp player chooses to "try" to win in a pub their score is "lower" then it should be and due to their high number of hours going to increase very slowly even on a winning streak.
Now I don't think there is a "solution" or really even a problem.... aside from the fact that if your playing against Schu you have a 50% chance of winning, and a 95% chance of feeling like your getting stomped the entire game.
Basically, (hive = skill * effort)... the effort specifically directed towards wining. Myself I'm an above average player and always want to win, even if it's due to a cheep tunnel and enzyme assisted CC rush(so my hive skill is 4K). So I think "I'm a badass". Then I see Schu stream and think, "damn, I too need some whiskey"
i play better when smurfing, averaging 10:1 ratio for kills and 40% accuracy, i actually only smurf while friends are around, good times.
is this because you're on rookie servers or is your smurf account seasoned?
btw, kind of unwise to go around with your name as 'ns2cheats.ru' or w/e your name was last i saw you in game, dont you think?
I played fair on the rookie server, one-two rounds then i moved on and left and then went to [TA] i only play, some good players once in a while on [TA] however most people are greens who have evolved, and what does "seasoned smurf account" even mean?
In regards to unwise, why does my name have any affect on you or anybody else?
By seasoned I mean that your smurf account is not in rookie status. You know, like seasons, time passing, "seasoned vet," etc. I thought my point would be clear by inferring that one would be allowed to play on rookie servers and one would not - the "seasoned" account.
Your name had no affect on me, in fact you had no affect on me during the round you joined me in either because you were IMMEDIATELY vote kicked because of the affect your stupid name had on the rest of the server. So I think it's unwise for you. That's all, was just sayin.
Your name had no affect on me, in fact you had no affect on me during the round you joined me in either because you were IMMEDIATELY vote kicked because of the affect your stupid name had on the rest of the server. So I think it's unwise for you. That's all, was just sayin.
bwahahahah. so people are now getting triggered by a sarcastic nickname? humanity is ripe for extinction...
Again, if this guy is smurfing or not, it doesn't really matter. It's a self-limiting problem in the end. Not completely, but largely.
A skilled smurf is a problem because they completely skew the shuffle and make their team more likely to win because that team's average hive skill should be higher, but isn't, right? A smurf is a problem if and only if they (and by extension their team) unduly benefits from a shuffle. The covertly amazing player with a bad hive skill value is paired with players of superior skill values, and the result is that the effective skill of the smurf's team is disproportionately higher than that of the opposing team.
If the player is consistently losing, they're not really creating big problems (at least not the sort of problems it's worth taking action over), even if you consider them a smurf for some other reason (because their aim is "too amazing for their skill level" or some other arbitrary assessment you're using).
If they're losing most and winning a few, winning some and losing some, or winning most and losing a few, it invites the question of how effectively they're "smurfing" in the first place.
If they're consistently winning, they're ranking up very quickly thanks to adagrad, which means that their skill progression probably looks something like this. That's the account of an obvious smurf. Sure, maybe that's not the true representation of this player's skill (yet), but it's getting there. Even though this player is smurfing, they can still lose, but the effect the losses have on their hive skill will be outpaced by the gain from their wins until they reach a natural equilibrium (somewhere 4k+, probably, for this player). And at that point they'll be expected to carry every time they're shuffled (which is its own problem for reasons mentioned elsewhere in this thread).
Maybe the final concern is that there's a period where they'll be smurfing too egregiously until their true hive skill is reached. In the very beginning, for example. But using the example provided, look how quickly they reached 2.8k skill. Now consider that most players never do. Also consider that once this player reaches 3k and begins the somewhat slower climb to 4k, anyone who still takes issue with their account will be quibbling over variances that I'm not sure any person is well-equipped to interpret at a glance (e.g., the absurdity of taking a stance like "this player is 3.2k and should clearly be 4k+ -- obvious smurf/cheater/hacker!").
So whatever. There are bigger fish to fry, and solid ways to help address this problem have already been outlined.
Your name had no affect on me, in fact you had no affect on me during the round you joined me in either because you were IMMEDIATELY vote kicked because of the affect your stupid name had on the rest of the server. So I think it's unwise for you. That's all, was just sayin.
bwahahahah. so people are now getting triggered by a sarcastic nickname? humanity is ripe for extinction...
yes, but still. if you tell a group of ppl who don't know you "hey I'm gonna cut you open" and walk around with a knife (even if that's not your intention), you can't blame ppl for reacting a certain way.
moultanoCreator of ns_shiva.Join Date: 2002-12-14Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
If anyone wants to do some reading, I've been trying to find references on whether using momentum with sparse gradients makes any sense. (By sparse I mean that most of the gradients are zero for most of the examples, because each game only has a small number of players in it.) To implement momentum precisely, you'd update every player on every game, which is both infeasible, and non-sensical. I think this would just degenerate into some constant factor change to the learning rate. It seems like it might work best if you defer the momentum update to the next time you have a nonzero gradient, but I'm having trouble finding evidence that anyone actually does that, and this seems like it should be a common issue when you have sparse features like words.
If anyone wants to do some reading, I've been trying to find references on whether using momentum with sparse gradients makes any sense. (By sparse I mean that most of the gradients are zero for most of the examples, because each game only has a small number of players in it.) To implement momentum precisely, you'd update every player on every game, which is both infeasible, and non-sensical. I think this would just degenerate into some constant factor change to the learning rate. It seems like it might work best if you defer the momentum update to the next time you have a nonzero gradient, but I'm having trouble finding evidence that anyone actually does that, and this seems like it should be a common issue when you have sparse features like words.
Short words, it's just a method to try to find faster a local max/min or hopefully a global max/min. The faster a skill rises, the more momentum it builds.
The behavior of the model/function mimics many things in the mechanical world which is well modeled and understood by control engineers.
For this to work reasonably well in the rank system (Im not sure where to best implement), you would probably need players playing close to 100% of their skill level and discretize lifeforms, otherwise the gradient/surface would become too noisy and the convergence unstable or too slow. And to test for stability is... tricky.
Comments
Sure, comp players don't always try as hard as they could. I also see players with extremely high mechanical skill with a lower hive skill than I would expect. These same players make very poor decisions. Wob has the highest hive skill by far, but I doubt he is the most gifted player to ever grace NS2. Wob is very skilled, but he just has the right combination of skills that allow him to win often in pubs.
I don't see how my simple statement is in any way a counter to that. I guess it was taken the wrong way.
I thought the people in the conversation were already likely aware of it, so took it as read. Fair enough if they were not.
Wob doesn't play many gathers and has barely played any competitive play for quite some time. He also tries really hard (nothing wrong with that). I've watched some other comp people (ADHD/Schu/asdf) for example. They gather and play a lot of comp, and generally don't try hard in pubs.
is this because you're on rookie servers or is your smurf account seasoned?
btw, kind of unwise to go around with your name as 'ns2cheats.ru' or w/e your name was last i saw you in game, dont you think?
i only smurf anymore when I'm tryin to log into main steam accnt and it asks for steamguard mobile authentication code and I'm not near my phone. i havent set up steamguard on my alt accnt so i dont need my phone glued to me to get into this one.
I played fair on the rookie server, one-two rounds then i moved on and left and then went to [TA] i only play, some good players once in a while on [TA] however most people are greens who have evolved, and what does "seasoned smurf account" even mean?
I think what i've realised is, if you just don't interact with people and enjoy the game it's far easier, especially when you have people targeting you even when you are smurfing, peoples names come to mind it's like they want to bitch about smurfs because they aren't even good at the game, even when you aren't stacking they still have a little cry...
Regardless ratio is probably closer to 9:1 though varies.
In regards to unwise, why does my name have any affect on you or anybody else? anyone that is going to stick around already knows whom eachother is anyway so it's not likely anyone else other then a few new people are going to have a moan and that is not the case, besides the server owners know who we are anyway so there's that. Hell i have a database of smurf accounts from the Australian community from when i was maintaining our servers, you'd be surprised at the amount of people who pretend like they've quit the game who are actually still around.
@Hobbeson as @Nordic had previously mentioned there is "mixing" what I was commenting on is that the "mixing" effect fails when people don't put for the the same effort on all servers. Specifically comp players who might not care as much about winning on Pub. Due to Hive ELO only looking at W/L the case competitive player base, hive score is lowered.
Granted that as long as the comp player puts forth a lower level of effort then hive score will still balance w/l. However, if a comp player chooses to "try" to win in a pub their score is "lower" then it should be and due to their high number of hours going to increase very slowly even on a winning streak.
Now I don't think there is a "solution" or really even a problem.... aside from the fact that if your playing against Schu you have a 50% chance of winning, and a 95% chance of feeling like your getting stomped the entire game.
Basically, (hive = skill * effort)... the effort specifically directed towards wining. Myself I'm an above average player and always want to win, even if it's due to a cheep tunnel and enzyme assisted CC rush(so my hive skill is 4K). So I think "I'm a badass". Then I see Schu stream and think, "damn, I too need some whiskey"
By seasoned I mean that your smurf account is not in rookie status. You know, like seasons, time passing, "seasoned vet," etc. I thought my point would be clear by inferring that one would be allowed to play on rookie servers and one would not - the "seasoned" account.
Your name had no affect on me, in fact you had no affect on me during the round you joined me in either because you were IMMEDIATELY vote kicked because of the affect your stupid name had on the rest of the server. So I think it's unwise for you. That's all, was just sayin.
bwahahahah. so people are now getting triggered by a sarcastic nickname? humanity is ripe for extinction...
A skilled smurf is a problem because they completely skew the shuffle and make their team more likely to win because that team's average hive skill should be higher, but isn't, right? A smurf is a problem if and only if they (and by extension their team) unduly benefits from a shuffle. The covertly amazing player with a bad hive skill value is paired with players of superior skill values, and the result is that the effective skill of the smurf's team is disproportionately higher than that of the opposing team.
- If the player is consistently losing, they're not really creating big problems (at least not the sort of problems it's worth taking action over), even if you consider them a smurf for some other reason (because their aim is "too amazing for their skill level" or some other arbitrary assessment you're using).
- If they're losing most and winning a few, winning some and losing some, or winning most and losing a few, it invites the question of how effectively they're "smurfing" in the first place.
- If they're consistently winning, they're ranking up very quickly thanks to adagrad, which means that their skill progression probably looks something like this. That's the account of an obvious smurf. Sure, maybe that's not the true representation of this player's skill (yet), but it's getting there. Even though this player is smurfing, they can still lose, but the effect the losses have on their hive skill will be outpaced by the gain from their wins until they reach a natural equilibrium (somewhere 4k+, probably, for this player). And at that point they'll be expected to carry every time they're shuffled (which is its own problem for reasons mentioned elsewhere in this thread).
- Maybe the final concern is that there's a period where they'll be smurfing too egregiously until their true hive skill is reached. In the very beginning, for example. But using the example provided, look how quickly they reached 2.8k skill. Now consider that most players never do. Also consider that once this player reaches 3k and begins the somewhat slower climb to 4k, anyone who still takes issue with their account will be quibbling over variances that I'm not sure any person is well-equipped to interpret at a glance (e.g., the absurdity of taking a stance like "this player is 3.2k and should clearly be 4k+ -- obvious smurf/cheater/hacker!").
So whatever. There are bigger fish to fry, and solid ways to help address this problem have already been outlined.Surely, you can connect those two quotes of yours there and see why maybe someone would have an issue, yea?
yes, but still. if you tell a group of ppl who don't know you "hey I'm gonna cut you open" and walk around with a knife (even if that's not your intention), you can't blame ppl for reacting a certain way.
If some kids are having problem / upset inside theirs little own virtual world ......
Im sure its not mine problem
http://distill.pub/2017/momentum/
Short words, it's just a method to try to find faster a local max/min or hopefully a global max/min. The faster a skill rises, the more momentum it builds.
The behavior of the model/function mimics many things in the mechanical world which is well modeled and understood by control engineers.
For this to work reasonably well in the rank system (Im not sure where to best implement), you would probably need players playing close to 100% of their skill level and discretize lifeforms, otherwise the gradient/surface would become too noisy and the convergence unstable or too slow. And to test for stability is... tricky.